ISLAB/CAISR - User contributions [en]

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2024-02-20T09:02:53Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles having spiral codes
|Prerequisites=Image Analysis, DEIS (Design of Embedded...)
|Supervisor=Josef Bigun
|Level=Master
|Status=Ongoing
}}
Build two driverless car-like robot vehicles scanning an "ice-rink", resurfacing every location by spiral codes.

Research challenges
Quantifying camera (localization) errors in stitching (textureless) ice images from different cameras
Obstacle detection on ice with camera on thin clients
Automatic parameter determination to construct scanning paths

Implementation Challenges:
Follow walls at precise distance, using proximity sensor.
Replicate following-walls behaviour with ceiling camera, and quantify error
Design simultaneous scanning patterns automatically for 1, and 2 robots.

The project is suggested by Bigsafe Technology AB.

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2024-02-19T12:32:02Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles having spiral codes
|Prerequisites=Image Analysis, DEIS (Design of Embedded...)
|Supervisor=Josef Bigun
|Level=Master
|Status=Ongoing
}}
Build two driverless car-like robot vehicles scanning an "ice-rink", resurfacing every location by spiral codes.

Research challenges
Quantifying camera (localization) errors in stitching (textureless) ice images from different cameras
Obstacle detection on ice with camera on thin clients
Automatic parameter determination to construct scanning paths

Implementation Challenges:
Follow walls at precise distance, using proximity sensor, based on time-of-flight.
Replicate following-walls behaviour with ceiling camera, and quantify error
Design simultaneous scanning patterns automatically for 1, and 2 robots.

The project is suggested by Bigsafe Technology AB.

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2024-02-19T12:30:56Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles having spiral codes
|Prerequisites=Image Analysis, DEIS (Design of Embedded...)
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}
Build two driverless car-like robot vehicles scanning an "ice-rink", resurfacing every location by spiral codes.

Research challenges
Quantifying camera (localization) errors in stitching (textureless) ice images from different cameras
Obstacle detection on ice with camera on thin clients
Automatic parameter determination to construct scanning paths

Implementation Challenges:
Follow walls at precise distance, using proximity sensor, based on time-of-flight.
Replicate following-walls behaviour with ceiling camera, and quantify error
Design simultaneous scanning patterns automatically for 1, and 2 robots.

The project is suggested by Bigsafe Technology AB.

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2023-10-12T20:44:30Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles having spiral codes
|Prerequisites=Image Analysis, DEIS (Design of Embedded...)
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}
Build two driverless car-like robot vehicles scanning an "ice-rink", resurfacing every location by spiral codes.

Research challenges
Quantifying camera localization errors against time-of-flight sensors
Obstacle detection on ice with camera on thin clients
Automatic parameter determination to construct scanning paths

Implementation Challenges:
Follow walls at precise distance, using proximity sensor, based on time-of-flight.
Replicate following-walls behaviour with ceiling camera, and quantify error
Design simultaneous scanning patterns automatically for 1, and 2 robots.

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2023-10-12T20:34:28Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles
|Prerequisites=Image Analysis, DEIS (Design of Embedded...)
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}
Build two driverless mini robot vehicles scanning an "ice-rink", resurfacing every location by spiral codes.

Challenges:
Follow walls at precise distance, using proximity sensor, based on time-of-flight.
Replicate following-walls behaviour with ceiling camera, and quantify error
Stop at obstacles, using onboard camera.
Design simultaneous scanning patterns automatically for 1, and 2 robots.

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2023-10-12T14:15:52Z

Josef:

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2023-10-12T14:14:20Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles
|Prerequisites=Image Analysis, DEIS
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2023-10-12T14:10:53Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}

Evaluation of captchas.

2022-10-31T10:58:11Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Evaluation of liveness detection provided by ocr captchas |Supervisor=Josef Bigun, Kevin Hernandez Diaz |Level=Master }} Intelligent vision s..."

{{StudentProjectTemplate
|Summary=Evaluation of liveness detection provided by ocr captchas
|Supervisor=Josef Bigun, Kevin Hernandez Diaz
|Level=Master
}}
Intelligent vision systems provide automatic
services in an increasing number of domains. One of the oldest is
automatic text recognition, also known as Optical Character
Recognition, OCR, was introduced in 1970'ies. For many people it became synonym with "scanning"
as it has been available as standard with most photocopy machines
since two decades and with widely available software, e.g. Adobe Acrobat.

Captchas, on the other hand have been relatively recently deployed as part of
automatic identity management systems in a variety of applications
such as automatic Schengen Visa appointment systems in
embassies, to make sure that valuable resources are not abused.
An important subset of captchas, called here ocr captchas, rely on recognition of
noise corrupted digit and letter sequences in the latin
alphabet. Here it is assumed that
humans are superior to machines to read them to the effect that such
captchas are in fact used to decide if someone who requests a valuable service is
effectively a human or a machine, also called liveness detection in
biometric identification. In the latter case, the system
automatically denies the requested service.

The thesis work will evaluate ocr captchas used in liveness detection
for web services. It will study how well modern CNN algorithms can be used
to detect purposely corrupted digit and letter sequences used in ocr
captchas, since such an algorithm will then be a threat to a valuable
service protected by captchas.

The problem is suggested by Bigsafe Technology AB.

Surface normal estimation by Spiral Codes

2021-10-11T11:30:50Z

Josef:

{{StudentProjectTemplate
|Summary=Estimating 3d surface normal from a single image
|References=J. Bigun "Vision with Direction", chapter 11, Springer, 2016.
|Prerequisites=Image Analysis, multi-variable calculus
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}
Estimating 3d surface normal from a single image is an ill defined problem.
However, if there is knowledge about the viewed pattern of a surface, the surface normal
can be estimated from (2d) local image orientation statistics.

The project will have the goal to review methods extracting surface orientation in 3d from 2D texture: shape from texture. Additionally, the question will be studied for a restricted class of patterns, in the form of spiral patterns.

Good knowledge in calculus and image analysis is required.

Surface normal estimation by Spiral Codes

2021-10-11T11:23:43Z

Josef:

Labeling by spiral codes for invariant recognition of garbage bags

2020-10-07T12:23:21Z

Josef: Blanked the page

A mechatronic system to recycle objects having Spiral Codes

2020-10-07T12:20:47Z

Josef: Blanked the page

Surface normal estimation by Spiral Codes

2019-10-19T21:10:50Z

Josef:

{{StudentProjectTemplate
|Summary=Estimating 3d surface normal from a single image
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}
Estimating 3d surface normal from a single image by using orientation statistics of a priori known patterns.

A mechatronic system to recycle objects having Spiral Codes

2019-10-19T21:09:02Z

Josef:

{{StudentProjectTemplate
|Summary=Mechatronic system to recycle objects
|Supervisor=Josef Bigun
|Level=Bachelor
|Status=Open
}}
Mechatronic system to recycle objects having Spiral Codes

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2019-10-19T21:01:10Z

Josef:

{{StudentProjectTemplate
|Summary=ice rink resurfacing system for selfdriving vehicles
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun
|Level=Master
|Status=Open
}}

Ice rink resurfacing system for selfdriving vehicles having spiral codes

2019-10-19T21:00:04Z

Josef: Created page with "{{StudentProjectTemplate |Summary= ice rink resurfacing system for selfdriving vehicles |Prerequisites=Image Analysis |Supervisor=Josef Bigun }}"

{{StudentProjectTemplate
|Summary= ice rink resurfacing system for selfdriving vehicles
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun
}}

A mechatronic system to recycle objects having Spiral Codes

2019-10-19T20:57:48Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Mechatronic system to recycle objects |Supervisor=Josef Bigun |Level=Bachelor }} Mechatronic system to recycle objects having Spiral Codes"

{{StudentProjectTemplate
|Summary=Mechatronic system to recycle objects
|Supervisor=Josef Bigun
|Level=Bachelor
}}
Mechatronic system to recycle objects having Spiral Codes

Surface normal estimation by Spiral Codes

2019-10-19T20:52:59Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Estimating 3d surface normal from a single image |Prerequisites=Image Analysis |Supervisor=Josef Bigun }} Estimating 3d surface normal from ..."

{{StudentProjectTemplate
|Summary=Estimating 3d surface normal from a single image
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun
}}
Estimating 3d surface normal from a single image by using orientation statistics of a priori known patterns.

Protecting bikers in traffic by computer vision

2019-09-27T13:21:58Z

Josef:

{{StudentProjectTemplate
|Summary=Protecting bikers in traffic by computer vision
|TimeFrame=HT2019
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun,
|Author=Josef Bigun
|Level=Master
|Status=Open
}}
Goal: To design and implement the visual processing of an early warning system to motor vehicles enabling biker protection in traffic.

The work includes automatic analysis of biker images with specially designed reflex patterns such that bikers, their motion intentions, and vehicles motion intentions will be discovered and communicated automatically earlier than current state of the art. The motion intentions will be conveyed by active messaging via spiral patterns and include signaling right/left turns of bikes (to vehicles) and relative position computations, braking of bikes including relative position computations, automatic messaging when intention conflicts arise with vehicle motions. The work includes computer vision software prototyping both on bike, and vehicle/infra-structure site, hardware prototyping of spirals including rotating spirals, but not other hardware proto-typing such as miniaturization, lighting at night, reflex version of labels, etc.

Human identification by handwriting of identity text

2019-09-27T13:19:35Z

Josef:

{{StudentProjectTemplate
|Summary=Identify a hand writer when repeated identity relevant text is available
|TimeFrame=HT2019
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez
|Level=Master
|Status=Open
|Title=Human identification by repeated hand-writing
}}
Goal: to extract identity relevant information from repeated instances of text entered by hand-written block-letters.

The project will work with the quiz scenario in mind, i.e. the participants are students who take quizes. A student will typically take several quizes and will enter the same or similar identity information by hand-writing on answer forms. The order of incoming response-forms is random, because at each quiz the students sit at different places, and not all take the same quizes. The number of persons taking the quiz is limited, e.g. the students of a school taking different courses.

The first mile-stone is to tell which quizes are hand-written by the same student. A second mile-stone is to extract the explicit identity of each student.

Digit recognition by lip-movements and time recursive Neural Networks

2019-09-27T13:15:28Z

Josef:

{{StudentProjectTemplate
|Summary=The project aims to recognize digits by lip movements and neural networks
|TimeFrame=HT2019
|Prerequisites=Good knowledge in Image Analysis and Computer Vision in 3D
|Supervisor=Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez,
|Level=Master
|Status=Open
}}
The project aims to use lip-movements to recognize digits visually only. This is interesting to verification of liveness in biometric identification as well as recognizing password utterences in public (=soundless speech). The work will study AI techniques in combination with Computer Vision techniqes, e.g. convolutional neural networks with short time memory and optical flow.

Liveness verification is an important issue in Biometrics. It corresponds to verifying that the signal , e.g. face-image, fingerprint, audio-speech, on which biometric recognition is based is authentic, coming from a live person, in contrast to a synthetic signal, including (dis)playing a video or speech from memory.

Also, audio utterances of words or digits when prompted for in public, such as trains, and busses, are not suitable to be used as passwords, for a variety of reasons. This includes overhearing by others, but also because speech is difficult to be used as biometric source, beside password information, for identification due to environment noise, e.g. vehicle noise. Accordingly, recognizing digits by lip-movements can be a way to mitigate the issues posed by speech in public, noisy, and crowded places.

Digit recognition by lip-movements and time recursive Neural Networks

2018-10-17T07:43:11Z

Josef:

{{StudentProjectTemplate
|Summary=The project aims to recognize digits by lip movements and neural networks
|Prerequisites=Good knowledge in Image Analysis and Computer Vision in 3D
|Supervisor=Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez,
|Level=Master
|Status=Open
}}
Liveness verification is an important issue in Biometrics. It corresponds to verifying that the signal , e.g. face-image, fingerprint, audio-speech, on which biometric recognition is based is authentic, coming from a live person, in contrast to a synthetic signal, including (dis)playing a video or speech from memory.

Also, audio utterances of words or digits when prompted for in public, such as trains, and busses, are not suitable to be used as passwords, for a variety of reasons. This includes overhearing by others, but also because speech is difficult to be used as biometric source, beside password information, for identification due to environment noise, e.g. vehicle noise. Accordingly, recognizing digits by lip-movements can be a way to mitigate the issues posed by speech in public, noisy, and crowded places.

The project aims to use lip-movements based digit recognition to contribute to verification of liveness as well as recognizing password utterences in public. It will be done by convolutional neural networks with short time memory and optical flow.

Digit recognition by lip-movements and time recursive Neural Networks

2018-10-17T07:31:18Z

Josef: Created page with "{{StudentProjectTemplate |Summary=The project aims to recognize digits by lip movements and neural networks |Prerequisites=Good knowledge in Image Analysis and Computer Vision..."

{{StudentProjectTemplate
|Summary=The project aims to recognize digits by lip movements and neural networks
|Prerequisites=Good knowledge in Image Analysis and Computer Vision in 3D
|Supervisor=Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez,
}}
Liveness verification is an important issue in Biometrics. It corresponds to verifying that the signal , e.g. face-image, fingerprint, audio-speech, on which biometric recognition is based is authentic, coming from a live person, in contrast to a synthetic signal, including (dis)playing a video or speech from memory.

Also, audio utterances of words or digits when prompted for in public, such as trains, and busses, are not suitable to be used as passwords, for a variety of reasons. This includes overhearing by others, but also because speech is difficult to be used as biometric source, beside password information, for identification due to environment noise, e.g. vehicle noise. Accordingly, recognizing digits by lip-movements can be a way to mitigate the issues posed by speech in public, noisy, and crowded places.

The project aims to use lip-movements based digit recognition to contribute to verification of liveness as well as recognizing password utterences in public. It will be done by convolutional neural networks with short time memory and optical flow.

Convolutional Neural Network (CNN) features behaviour in the context of textures

2018-10-15T20:28:09Z

Josef:

{{StudentProjectTemplate
|Summary=The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.
|Prerequisites=Image Analysis: Grade 4 or 5
|Supervisor=Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez.
|Level=Master
|Status=Open
}}
First it will Investigate how the cnn's behave when the number of classes increase, potentially to infinity. Here, syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises finding the local orientations on a series of Frequency Modulated test images and associated certainties. The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test type images this equals to their sizes).

Second, Convolutional Neural Network (CNN) features discrimination power will be be quantized in terms of within and between classsvariances, with different textures in a mosaic of texture images, cut from real aerial images and Brodatz textures.

Convolutional Neural Network (CNN) responses when the number of classes increase

2018-10-15T20:21:54Z

Josef:

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) features behaviour in the context of textures
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Level=Master
|Status=Internal Draft
}}
Convolutional Neural Network (CNN) features behaviour in the context of textures

The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.

Image analysis

Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez.

The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.

Master, Open

First it will Investigate how the cnn's behave when the number of classes increase, potentially to infinity. Here, syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises finding the local orientations on a series of Frequency Modulated test images and associated certainties. The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Second, Convolutional Neural Network (CNN) features discrimination power will be be quantized in terms of within and between classsvariances, with different textures in a mosaic of texture images, cut from real aerial images and Brodatz textures.

Labeling by spiral codes for invariant recognition of garbage bags

2018-10-15T20:00:10Z

Josef:

{{StudentProjectTemplate
|Summary=Labeling by spiral codes for invariant recognition of garbage bags
|Prerequisites=Image Analysis: Grade 4 or 5
|Supervisor=Josef Bigun,
|Author=Josef Bigun
|Level=Master
|Status=Open
}}
Goal: To design spiral codes and recognition method of different fractions of garbage bags at arbitrary orientation on conveyor belt of recycling station

The work includes synthesis of spiral labels for 4 to 8 fractions of recyclable materials sorted into plastic bags by households. Here fraction means a recyclable material. For example food remnants, plastics, cartons, metal, colored glass, and uncolored glass are different fractions. The purpose is to increase gain of material via recycling, reducing the environmental footprint. The key element of the environmental contribution is to make recycling easy for households as the number of fractions increases while this should not increase the (human) work at garbage collection trucks and recycling stations. Households, will thus use differently labeled bags for different fractions and put filled bags into the same garbage container. Currently, color is used as label, limiting the number of fractions and performance of the correct recognition in comparison with shape based labels. In some municipalities the household containers have different compartments for each fraction but this increases the complexity of collection and transportation, resulting in larger footprint on the environment. The work includes bag pattern prototyping, data collection, and computer vision software prototyping. The project is suggested by Bigsafe Technology AB.

Labeling by spiral codes for invariant recognition of garbage bags

2018-10-15T19:56:14Z

Josef:

{{StudentProjectTemplate
|Summary=Labeling by spiral codes for invariant recognition of garbage bags
|References=
|Prerequisites=Image Analysis: Grade 4 or 5
|Supervisor=Josef Bigun,
|Author=Josef Bigun
|Level=Master
|Status=Open
}}
Goal: To design spiral codes and recognition method of different fractions of garbage bags at arbitrary orientation on conveyor belt of recycling station

The work includes synthesis of spiral labels for 4 to 8 fractions of recyclable materials sorted into plastic bags by housholds. Here fraction means a recyclable material. For example food remnants, plastics, cartons, metal, colored glas, and uncolored glas are different fractions. The purpose is to increase gain of material via recycling, reducing the environmental footprint. The key element of the environmental contribution is to make recycling easy for households as the number of fractions increases while this should not increase the (human) work at garbage collection trucks and recycling stations. Households, will thus use differently labeled bags for each fraction and put filled bags into the same garbage container. Currently, color is used as label, limitting the number of fractions and performance of the correct recognition in comparison with shape based labels. In some municipalities the household containers have different compartments for each fraction but this increases the complexity of the collection and transportation, having bigger footprints on the environment. The work includes bag pattern prototyping, data collection, and computer vision software prototyping. The project is suggested by Bigsafe Technology AB.

Labeling by spiral codes for invariant recognition of garbage bags

2018-10-15T19:55:13Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Labeling by spiral codes for invariant recognition of garbage bags |Supervisor=Josef Bigun }}"

{{StudentProjectTemplate
|Summary=Labeling by spiral codes for invariant recognition of garbage bags
|Supervisor=Josef Bigun
}}

Human identification by handwriting of identity text

2018-10-15T14:54:30Z

Josef:

{{StudentProjectTemplate
|Summary=Identify a hand writer when repeated identity relevant text is available
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez
|Level=Master
|Status=Open
|Title=Human identification by repeated hand-writing
}}
Goal: to extract identity relevant information from repeated instances of text entered by hand-writting automatically using image processing.

The project will work with the quiz scenario in mind, i.e. the participants are students who take quizes. A student will typically take several quizes and will enter the same or similar identity information by hand-writing on answer forms. The order of incoming response-forms is random, because at each quiz the students sit at different places, and not all take the same quizes. The number of persons taking the quiz is limited, e.g. the students of a school taking different courses.

The first mile-stone is to tell which quizes are hand-written by the same student. A second mile-stone is to extract the explicit identity of each student.

Human identification by handwriting of identity text

2018-10-15T14:53:33Z

Josef:

{{StudentProjectTemplate
|Summary=Identify a hand writer when repeated identity relevant text is available
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun,
|Level=Master
|Status=Open
|Title=Human identification by repeated hand-writing
}}
Goal: to extract identity relevant information from repeated instances of text entered by hand-writting automatically using image processing.

The project will work with the quiz scenario in mind, i.e. the participants are students who take quizes. A student will typically take several quizes and will enter the same or similar identity information by hand-writing on answer forms. The order of incoming response-forms is random, because at each quiz the students sit at different places, and not all take the same quizes. The number of persons taking the quiz is limited, e.g. the students of a school taking different courses.

The first mile-stone is to tell which quizes are hand-written by the same student. A second mile-stone is to extract the explicit identity of each student.

Convolutional Neural Network (CNN) features behaviour in the context of textures

2018-10-15T14:51:21Z

Josef:

Convolutional Neural Network (CNN) features behaviour in the context of textures

2018-10-15T14:50:46Z

Josef:

Convolutional Neural Network (CNN) features behaviour in the context of textures

2018-10-15T14:48:40Z

Josef: Created page with "{{StudentProjectTemplate |Summary=The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures. |Prerequisites=Image A..."

{{StudentProjectTemplate
|Summary=The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez.
|Level=Master
|Status=Open
}}
The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.

First it will Investigate how the cnn's behave when the number of classes increase, potentially to infinity. Here, syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises finding the local orientations on a series of Frequency Modulated test images and associated certainties. The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Second, Convolutional Neural Network (CNN) features discrimination power will be be quantized in terms of within and between classsvariances, with different textures in a mosaic of texture images, cut from real aerial images and Brodatz textures.

Convolutional Neural Network (CNN) responses when the number of classes increase

2018-10-15T14:41:52Z

Josef:

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) features behaviour in the context of textures
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Level=Master
|Status=Open
}}
Convolutional Neural Network (CNN) features behaviour in the context of textures

The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.

Image analysis

Josef Bigun, Kevin Hernandez-Diaz, Fernando Alonso-Fernandez.

The project aims to quantify the behaviour of Convolutional Neural Network (CNN) features in the context of textures.

Master, Open

First it will Investigate how the cnn's behave when the number of classes increase, potentially to infinity. Here, syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises finding the local orientations on a series of Frequency Modulated test images and associated certainties. The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Second, Convolutional Neural Network (CNN) features discrimination power will be be quantized in terms of within and between classsvariances, with different textures in a mosaic of texture images, cut from real aerial images and Brodatz textures.

Convolutional Neural Network (CNN) responses when the number of classes increase

2018-10-15T14:34:14Z

Josef:

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) features behaviour in the context of textures
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Level=Master
|Status=Open
}}
Investigate how the cnn's behave when the number of classes increase, potentially to infinity.

Syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises finding the local orientations on a series of Frequency Modulated test images and associated certainties. The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

RAQUEL Robot Assisted QUiz Espying of Learners

2017-10-13T12:06:54Z

Josef:

{{StudentProjectTemplate
|Title=RAQUEL Robot Assisted QUiz Espying Learners
|Summary=RAQUEL Robot Assisted QUiz Espying Learners
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Martin Cooney, Fernando Alonso Fernandez
|Level=Master
|Status=Open
}}
Robot Assisted QUiz Labeling by face recognition.
We teach baxter to be a quiz assistant.

RAQUEL Robot Assisted QUiz Espying of Learners

2017-10-13T12:04:53Z

Josef:

{{StudentProjectTemplate
|Summary=RAQUEL Robot Assisted QUiz Espying Learners
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Martin Cooney, Fernando Alonso Fernandez
|Level=Master
|Status=Open
}}
Robot Assisted QUiz Labeling by face recognition.
We teach baxter to be a quiz assistant.

RAQUEL Robot Assisted QUiz Espying of Learners

2017-10-12T05:31:25Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Create StudentProjectForm: RAQUEL Robot Assisted QUiz Espying of LearnersCreate StudentProjectForm: RAQUEL Robot Assisted QUiz Labeling by f..."

{{StudentProjectTemplate
|Summary=Create StudentProjectForm: RAQUEL Robot Assisted QUiz Espying of LearnersCreate StudentProjectForm: RAQUEL Robot Assisted QUiz Labeling by face recognition
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun, Martin Cooney, Fernando Alonso Fernandez
}}
Robot Assisted QUiz Labeling by face recognition.
We teach baxter to be a quiz assistant.

Convolutional Neural Network (CNN) responses when the number of classes increase

2017-10-05T10:40:38Z

Josef:

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) responses when the number of classes increase
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Level=Master
|Status=Open
}}
Investigate how the cnn's behave when the number of classes increase, potentially to infinity.

Syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises
finding the local orientations on a series of Frequency Modulated test images and associated certainties.

A similar problem is estimating the Local Frequency.

The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Convolutional Neural Network (CNN) responses when the number of classes increase

2017-10-05T10:39:59Z

Josef:

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) responses when the number of classes increase
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Author=
|Level=Master
|Status=Ongoing
}}
Investigate how the cnn's behave when the number of classes increase, potentially to infinity.

Syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises
finding the local orientations on a series of Frequency Modulated test images and associated certainties.

A similar problem is estimating the Local Frequency.

The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Convolutional Neural Network (CNN) responses when the number of classes increase

2017-09-26T08:23:19Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Convolutional Neural Network (CNN) responses when the number of classes increase |Prerequisites=image analysis |Supervisor=Josef Bigun, Fern..."

{{StudentProjectTemplate
|Summary=Convolutional Neural Network (CNN) responses when the number of classes increase
|Prerequisites=image analysis
|Supervisor=Josef Bigun, Fernando Alonso-Fernandez,
|Author=Alexander Galozy
}}
Investigate how the cnn's behave when the number of classes increase, potentially to infinity.

Syntethic images can be used since the groundtruth is known. One such classificaiton problem comprises
finding the local orientations on a series of Frequency Modulated test images and associated certainties.

A similar problem is estimating the Local Frequency.

The number of classes can be made as large as one wishes in both cases, the
discretizaton grid being the limit (In FM-test images this equals to their sizes).

Chess playing humanoid robot by vision

2017-09-26T07:42:11Z

Josef:

{{StudentProjectTemplate
|Summary=Chess playing humanoid robot by vision
|References=https://www.youtube.com/watch?v=gXOkWuSCkRI
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun,
|Author=Joseph T. Sachin
|Level=Master
|Status=Ongoing
}}
Goal: to design and implement the visual processing of a humanoid robot enabling it to play chess with real chess-pieces.

The work includes mapping images of real chess pieces on a chess-board to chess-maps, as well as to move chess pieces to new locations so that an external software can lead the chess-logics. A design choice for the latter is included but not its development.

Protecting bikers in traffic by computer vision

2016-11-07T13:48:22Z

Josef:

{{StudentProjectTemplate
|Summary=Protecting bikers in traffic by computer vision
|References=
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun,
|Author=Josef Bigun
|Level=Master
|Status=Open
}}
Goal: To design and implement the visual processing of an early warning system to motor vehicles enabling biker protection in traffic.

The work includes automatic analysis of biker images with specially designed reflex patterns such that bikers, their motion intentions, and vehicles motion intentions will be discovered and communicated automatically earlier than current state of the art. The motion intentions will be conveyed by active messaging via spiral patterns and include signaling right/left turns of bikes (to vehicles) and relative position computations, braking of bikes including relative position computations, automatic messaging when intention conflicts arise with vehicle motions. The work includes computer vision software prototyping both on bike, and vehicle/infra-structure site, hardware prototyping of spirals including rotating spirals, but not other hardware proto-typing such as miniaturization, lighting at night, reflex version of labels, etc.

Protecting bikers in traffic by computer vision

2016-11-07T13:47:39Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Protecting bikers in traffic by computer vision |Supervisor=Josef Bigun, }}"

{{StudentProjectTemplate
|Summary=Protecting bikers in traffic by computer vision
|Supervisor=Josef Bigun,
}}

Human identification by handwriting of identity text

2016-11-07T13:35:34Z

Josef:

{{StudentProjectTemplate
|Title=Human identification by repeated hand-writing
|Summary=Identify a hand writer when repeated identity relevant text is available
|Supervisor=Josef Bigun,
|Level=Master
|Status=Open
}}

Goal: to extract identity relevant information from repeated instances of text entered by hand-writting automatically using image processing.

The project will work with the quiz scenario in mind, i.e. the participants are students who take quizes. A student will typically take several quizes and will enter the same or similar identity information by hand-writing on answer forms. The order of incoming response-forms is random, because at each quiz the students sit at different places, and not all take the same quizes. The number of persons taking the quiz is limited, e.g. the students of a school taking different courses.

The first mile-stone is to tell which quizes are hand-written by the same student. A second mile-stone is to extract the explicit identity of each student.

Name of the Cnew project

2016-11-07T10:45:16Z

Josef: Josef moved page Name of the Cnew project to Chess playing humanoid robot by vision

#REDIRECT [[Chess playing humanoid robot by vision]]

Chess playing humanoid robot by vision

2016-11-07T10:45:16Z

Josef: Josef moved page Name of the Cnew project to Chess playing humanoid robot by vision

{{StudentProjectTemplate
|Summary=Chess playing humanoid robot by vision
|References=https://www.youtube.com/watch?v=gXOkWuSCkRI
|Prerequisites=Image Analysis
|Supervisor=Josef Bigun,
|Author=Josef Bigun
|Level=Master
|Status=Open
}}
Goal: to design and implement the visual processing of a humanoid robot enabling it to play chess with real chess-pieces.

The work includes mapping images of real chess pieces on a chess-board to chess-maps, as well as to move chess pieces to new locations so that an external software can lead the chess-logics. A design choice for the latter is included but not its development.

Chess playing humanoid robots by vision

2016-11-07T10:38:16Z

Josef:

{{StudentProjectTemplate
|Summary=Chess playing humanoid robot by vision
|Supervisor=Josef Bigun
}}
Goal: to design and implement the visual processing of a humanoid robot enabling it to play chess with real chess-pieces.

The work includes mapping images of real chess pieces on a chess-board to chess-maps, as well as to move chess pieces to new locations so that an external software can lead the chess-logics. A design choice for the latter is included but not its development.

Chess playing humanoid robots by vision

2016-11-07T10:35:54Z

Josef: Created page with "{{StudentProjectTemplate |Summary=Chess playing humanoid robot by vision |Supervisor=Josef Bigun, }} Goal: to design and implement the visual processing of a humanoid robot ..."

{{StudentProjectTemplate
|Summary=Chess playing humanoid robot by vision
|Supervisor=Josef Bigun,
}}
Goal: to design and implement the visual processing of a humanoid robot enabling it to play chess with real chess-pieces.

The work includes mapping images of real chess pieces on a chess-board to chess-maps, as well as to move chess pieces to new locations so that an external software can lead the chess-logics. A design choice for the latter is included but not its development.