| Description |
Data |
Model |
Starred |
| • VinMap, a dataset of Vietnam maps containing Truong Sa and Hoang Sa as well as maps not containing those islands |
✓ |
✓ |
...★ |
|
|
| • EgoPer, a method for error detection on egocentric procedural task videos |
✓ |
✓ |
...★ |
|
|
| • Blur2Blur, an image debluring image that works by converts a blurry image from one blurry domain to another |
|
✓ |
...★ |
|
|
| • HAT, Human Attention Transformer for unifying top-down and bottom-up scanpath prediction |
|
✓ |
...★ |
|
|
| • Diverse and Wearable Counting (DWC), a method and a dataset for counting human actions of interest using sensor data from wearable devices, allowing users to provide exemplars of the actions they want to count by vocalizing pre-defined sounds "one", "two", and "three". |
✓ |
✓ |
...★ |
|
|
| • ICACount. A GUI and algorithm for interactive counting |
|
✓ |
...★ |
|
|
| • HyperCut and RB2V. HyperCut is an algorithm for image-to-video deblurring. RB2V is a real blur2vid dataset captured by a hardware setup with two GoPro cameras and a beam splitter. RB2V contains three subsets: RB2V-Face, RB2V-Hand, and RB2V-Street. |
✓ |
✓ |
...★ |
|
|
| • Scenses100, a model and a dataset for self-supervised scence-adaptive object detection. The dataset contains 100 videos from stationary cameras, each with a duration of several hours. |
✓ |
✓ |
...★ |
|
|
| • GazeFormer, a method that can predict the fixations that a person makes while searching for a target (e.g., a fork) despite the unavailability of search fixations (e.g., a person looking for a fork) for model training. |
|
✓ |
...★ |
|
|
| • Counting-DTR + FSCD-147, a model and a dataset for few-shot counting and detection |
✓ |
✓ |
...★ |
|
|
| • FamNet + FSC-147, a model and a dataset for few-shot object counting |
✓ |
✓ |
...★ |
|
|
| • TopoCount, a method that uses topological constraints for crowd counting and people localization |
|
✓ |
...★ |
|
|
| • DM-Count, a method for crowd counting based on Optimal Transport loss |
|
✓ |
...★ |
|
|
| • Exemplar-free counting, a method for counting objects in images without exemplars |
|
✓ |
...★ |
|
|
| • Self-supervised Point Clound Analysis, a method for point cloud analysis based on self-supervised learning with multi-view rendering |
|
✓ |
...★ |
|
|
| • CrossKD, a method for cross modality retrieval based on knowledge distillation |
|
✓ |
...★ |
|
|
| • Target-absent human attention, a method for predicting gaze behavior searching for a target that is not in a scene |
|
✓ |
...★ |
|
|
| • VinText, a daset for Vietnamese scene text recognition, and an algorithmn for scene text recognition that laverages the dictionary in both training and inference time |
✓ |
✓ |
...★ |
|
|
| • MagMet, a method for progressive semantic segmentation |
|
✓ |
...★ |
|
|
| • Blur Kernel Space, a method for image debluring with encoded blur kernel space |
|
✓ |
...★ |
|
|
| • CPM, a method and a datset for makeup transfer |
✓ |
✓ |
...★ |
|
|
| • IRL Scanpath prediction, a method for predicting scanpath based on inverse reinforcement learning |
|
✓ |
...★ |
|
|
| • Image foveation, a python library for doing image foveation |
|
✓ |
...★ |
|
|
| • Coco-Search18, a laboratory-quality dataset of goal-directed behavior large enough to train deep-network models. It consists of the eye gaze behavior from 10 people searching for each of 18 target-object categories in 6202 natural-scene images, yielding about 300,000 search fixations. |
✓ |
|
|
|
|
| • Coco-FreeView, a laboratory-quality dataset of free viewing behavior. It contains the same natural images used in COCO-Search18, but labeled with 822,602 eye fixations from a free-viewing task. |
✓ |
|
|
|
|
| • Microwave-Clock, eye gaze dataset of search behavior for two search targes: Microwave and Clocks. The dataset contains 2183 images, each is viewed by at least 30 human participants. |
✓ |
|
|
|
|
| • Crowd Counting Gaze, a dataset containing gaze behavior of 10 human participants who are tasked with counting the number of people in 30 crowd images |
✓ |
|
|
|
|
| • HandLer + YoutubeHands, a method and dataset for multiple hand tracking |
✓ |
✓ |
...★ |
|
|
| • BodyHands, a method and a datset for hand detection and hand body association |
✓ |
✓ |
...★ |
|
|
| • ContactHands, a method and a dataset for detecting hands and estimating the physical contacts of hands |
✓ |
✓ |
...★ |
|
|
| • Hand-CNN, Keras code and model for hand detection for images in the wild |
|
✓ |
...★ |
|
|
| • TV-Hand, Coco-Hand, two datasets containing in-the-wild images with hand location annotation |
✓ |
|
|
|
|
| • WorkingHands, a dataset of hands interacting with hand-held tools, seen from a overhead camera obersving a table-top working area. This contains both synthetic and real images with segmentation masks |
✓ |
|
|
|
|
| • SBU shadow dataset |
✓ |
|
|
|
|
| • Gif2video, a method to convert a gif image into a video |
|
✓ |
...★ |
|
|
| • StockEmotion, a large-scale dataset of human emotion, a pre-trained model and code for extracting emotion representation vector for an image |
✓ |
✓ |
...★ |
|
|
| • MHLfuncs, collection of Matlab functions that are frequently used in my research, which might also be useful for some other people. The functions support various tasks, including: classification (e.g., Kernel ridge regression, kernel computation, kNN), evaluation (e.g., computing Average precision, plotting precision recall curve, plotting ROC curve), optimization (e.g., Sgd, Asgd), k-means (e.g., parallel k-means, permuation k-means), Misc. utility functions (e.g., create latex table, progress bar, compute square distance), Video processing (frame extraction, mp4 creation, shot boundary detection, video thread linking). Version 4.1, last updated 14 Jul 2015. |
|
✓ |
...★ |
|
|
| • Upper body detector, Matlab code for detecting upper bodies of people in video frames. This code is based on deformable part model, also leveraing the prototypical configrations of upper bodies in TV series for better detection. Version 1.0, last updated 22 Apr 14 |
|
✓ |
|
|
|
| • Discriminative Sub-categorization, version 1.0, last updated 29 Apr 13 |
|
✓ |
|
|
|
| • Max-Margin Early Event Detectors, version 2.0, last updated 1 Mar 2016 |
|
✓ |
|
|
|
| • Joint localization and classification, version 4.0, last updated 8 Jul 11. |
|
✓ |
|
|
|
| • Branch-and-bound and dynamic programming code, a set of efficient algorithms for optimal localization of bounding box and temporal segments, based on branch and bound and dynamic programming algorithms. Version 2.0, last updated 8 Jul 2011. |
|
✓ |
|
|
|
| • Robust Kernel PCA, implementation of a robust kernel principal component analysis algorithm, and demo code for its applications to handle noise, occlusion, and missing data in images. Version 1.0 |
|
✓ |
|
|
|
| • Feature selection for SVM, version 1.0 |
|
✓ |
|
|
|