Jan van Gemert

Contact

Image of Jan van Gemert

Dr. Jan van Gemert
Head of the Computer vision lab
Faculty of Electrical Engineering, Mathematics and Computer Science
Delft University of Technology

Visit: Van Mourik Broekmanweg 6, Delft (building 28), 5th floor (East) room E5.080
Post: Postbus 5031, 2600 GA Delft.
Phone: (+31)(0)15 278 84 34
Email: J.C.vanGemert at sign TUDelft dot nl
github: https://github.com/jvgemert

Recent news (2018):

- Accepted paper. TIP 2018, "Divide and Count: Generic Object Counting by Image Divisions" (pdf).
- Accepted paper. ICIP 2018, "Recurrent knowledge distillation" (pdf).
- Award. ECCV 2018 workshop best paper award for "Hand-tremor frequency estimation in videos" (pdf).
- Workshop. Organizing ECCV'18 workshop Anticipating Human Behavior.
- Accepted paper. ECCV 2018 workshop, "Using phase instead of optical flow for action recognition" (pdf).
- Accepted paper. ECCV 2018 workshop, "Hand-tremor frequency estimation in videos" (pdf).
- Accepted paper. PRL 2018, "Asymmetric kernel in Gaussian Processes for learning target variance" (pdf).
- Accepted paper. International Archives of the Photogrammetry, Remote Sensing & Spatial Information Sciences, "Shape based classification of seismic building structural types".
- Speaker. Keynote speaker at the AI expo Europe (presentation slides: Deep Learning Bytes).
- Seminar. Organizing the "Delft deep learning seminar" (Flyer).
- Seminar. Taught the "Deep Learning Workshop" for TU Delft personell.
- Guest lecture at the iQ winterschool 2018 on Machine Learning Applied to Quantitative Analysis of Medical Images.
- Interview. Interviewed for Humans of TU Delft.
- Visiting researcher. I am visiting the Computer Vision Center in Barcelona for six weeks.
- Grant. Obtained a Take-off grant for visually inspecting aircraft engines.

News (2017):

- Grant. Obtained funding through a Perpectief grant on EDL: Efficient Deep Learning.
- Grant. Obtained funding through a Perpectief grant on Citius Altius Sanius: Injury-free exercise for everyone.
- Youtube video of presentation available. For the ICCV 2017 paper "Active Decision Boundary Annotation with Deep Generative Models" (pdf) the video of the spotlight is online.
- Code available. For the ICCV 2017 paper "Active Decision Boundary Annotation with Deep Generative Models" (pdf) we have python code available on github.
- Award. BMVC 2017: Received Outstanding reviewer award.
- Press release. Nieuwe aanpak bewegingsanalyse blijkt nuttig voor Parkinson, vliegtuigmotoren en sport.
- Accepted paper. ICCV 2017, "Active Decision Boundary Annotation with Deep Generative Models" (pdf).
- Accepted paper. BMVC 2017, "Object-Extent Pooling for Weakly Supervised Single-Shot Localization" (pdf).
- Grant. Obtained funding for a PhD student from NWO Commit2Data for analyzing videos over time.
- Award. CVPR 2017: Received Outstanding reviewer award.
- Accepted paper. VAST 2017, "DeepEyes: Progressive Visual Analytics for Designing Deep Neural Networks" (pdf).
- Accepted paper. ICIAP 2017, "One-Step Time-Dependent Future Video Frame Prediction with a Convolutional Encoder-Decoder Neural Network".
- Accepted paper. CVPR 2017, "Video Acceleration Magnification" (pdf | Project).
- Accepted paper. DLM 2017, "Vision-based Detection of Acoustic Timed Events: a Case Study on Clarinet Note Onsets" (pdf, video).
- Accepted paper. IJCV 2017, "Tubelets: Unsupervised action proposals from spatiotemporal super-voxels" (pdf).
- Accepted paper. TIP 2017, "Con-Text: Text Detection for Fine-grained Object Classification". To appear.
- Workshop. Organizing CVPR'17 workshop Brave new ideas for motion representations in video.
- Interview. In medical delta about the Technology In Motion (TIM) lab.
- Postdoc position available. ArchiMediaL: deep learning and knowledge representations (academic transfer).
- Grant. Obtained funding for a postdoc from Volkswagen Stiftung for visually matching various architectural modalities. Project ArchiMediaL together with Prof. Carola Hein from Architecture @ TU-Delft, and Victor de Boer from Knowledge representations @ VU.
- Grant. Obtained a pilot grant from Sports engineering institute for visually analyzing exercising in the gym.
- Grant. Obtained a Take-off grant for visually analyzing recycling plastic.

News (2016):

- Conference. Organizing the Netherlands Conference on Computer Vision (NCCV'16)
- Workshop. Organized ECCV'16 workshop on Brave new ideas for motion representations in video.
- Conference. Co-Organized the European Conference on Computer Vision (ECCV'16)
- Data available. Code of the ICCV 2015 paper "Objects2action: Classifying and localizing actions without any video example" (pdf). All the embeddings we used are available.
- Accepted paper. ECCV 2016 workshop, "Making a Case for Learning Motion Representations with Phase" [pdf]
- Code available. of the CVPR 2016 paper "Structured Receptive Fields in CNNs" (pdf). Caffe and Theano code availabe on github.
- Accepted paper. ECCV 2016, "Spot On: Action Localization from Pointly-Supervised Proposals" [pdf]. Hollywood2Tubes dataset.
- Accepted paper. ECCV 2016, "Depth-aware Motion Magnification" [pdf, Supplemental]
- Organizing workshop. ECCV'16 Workshop on: Brave new ideas for motion representations in videos
- Accepted paper. CVIU 2016, "No Spare Parts: Sharing Part Detectors for Image Categorization"
- Award. Amazon Picking Challenge 2016: The team I advised on deep learning won two tasks. See news articles in The Verge and engadget.
- Award. CVPR 2016: Received Outstanding reviewer award.
- Accepted paper. ICIP 2016, "Featureless: Bypassing Feature Extraction in Action Categorization" (pdf)
- Accepted paper. CVIU 2016, "Large Scale Gaussian Process for Overlap-based Object Proposal Scoring" (pdf)
- Accepted paper. CVPR 2016, "Structured Receptive Fields in CNNs" (pdf)
- Presentation available. Of the oral presentation at the Kivi Kooy symposium of "De plenoptische mens: alle visuele stromen bevaren" (Presentation)
- Poster available. Of the presentation at Delft Data Science and sports, "Computer Vision for Sports" (Poster)
- Presentation available. Of the oral presentation at BMVC 2015 of "APT: Action localization Proposals from dense Trajectories" (Presentation)

News (2015):

- Code available. Of "Phase Based Video Motion Processing" as implemented during the Lorentz workshop "ICT with Industry 2015". Python source code available on github
- Data available. Of "What do 15,000 Object Categories Tell Us About Classifying and Localizing Actions?" (.pdf) features available
- Code available. Of "APT: Action localization Proposals from dense Trajectories" (.pdf) python code and pre-computed tubes available on github
- Accepted paper. ICCV 2015, "Objects2action: Classifying and localizing actions without any video example" (pdf)
- Accepted paper. NCCV 2015, "BING3D: Fast Spatio-Temporal Proposals for Action Localization" (pdf)
- Accepted paper. MMM 2016, "Exploring the Long Tail of Social Media Tags" (pdf)
- Accepted paper. BMVC 2015, oral presentation, "APT: Action localization Proposals from dense Trajectories" (.pdf)
- New position: Assistant professor at Delft University of Technology.
- Supervision. Martine van den Berg, graduated on "Mobile Object Recognition: What are the Minimum Criteria that are Acceptable to Museum Visitors?"
- Supervision. Johan Sundin MSc, graduated on "Image Editing from multiple photos to one single best" (pdf)
- Supervision. June, supervising 8 profile project MSc AI students
- Accepted paper. CVPR 2015, "What do 15,000 Object Categories Tell Us About Classifying and Localizing Actions?" (pdf)
- Award. CVPR 2015: Received Outstanding reviewer award
- Accepted paper. ICMR 2015, "Bag-of-Fragments: Selecting and encoding video fragments for event detection and recounting" (pdf)
- Accepted paper. ICIP 2015, "Per-patch metric learning for robust image matching" (pdf)

Valid XHTML 1.0 Transitional
Valid CSS!