Chris Pollett > Students >
Matharu

    ( Print View)

    [Bio]

    [Blog]

    [CS 297 Proposal]

    [Deliverable 1]

    [Rapid Object Detection using a Boosted Cascade of Simple Features-PDF]

    [Deliverable 2]

    [Deliverable 3-PDF]

    [Deliverable 4]

    [CS297 Report-PDF]

    [CS298 Proposal]

    [CS298 Report-PDF]

    [CS298 Defense Slides-PDF]

Project Blog

CS 298

Week 16: Mar 18, 2021

Minutes

  • Defense Scheduled and Approved

Week 15: May 11, 2021

Minutes

  • Review of CS298 Defense Slides

TODO:

  • Incorporate the changes suggested

Week 14: May 4, 2021

Minutes

  • Review of CS298 Report

TODO:

  • Incorporate the changes suggested

Week 13: Apr 27, 2021

Minutes

  • Showed the implementation of the Attention Model
  • Showed the final implementation of the whole model (Encoder + Decoder): Action Classifier and Heatmap Decoder

TODO:

  • Change the layers in Attention Model and perform experiments

Week 12: Apr 20, 2021

Minutes

  • Showed the design of added Attention Model in Encoder

TODO:

  • Complete the whole architecture of the model
  • Perform the changes suggested in the CS298 Draft Report

Week 11: Apr 13, 2021

Minutes

  • Discuss the problem of poor model learning

TODO:

  • Incorporate Attention Model to solve the model performance
  • Start the first draft of Report

Week 10: Apr 6, 2021

Minutes

  • Showed the implementation of Heatmap Predictor
  • Discussed the error of noise in results of Heatmap

TODO:

  • Fix the error in the heatmap model by changing its architecture

Week 9: Mar 30, 2021

Minutes

  • Discussed the architecture of Fully convolutional neural network with Transposed Convolution layer for heatmap predictor

TODO:

  • Implement the discussed architecture

Week 8: Mar 23, 2021

Minutes

  • Showed and discussed the plausible design of classification model(Parallel Encoder + ConvLSTM)

TODO:

  • Fix the error in the classification model and implement it

Week 7: Mar 16, 2021

Minutes

  • Showed the implementation of CNN and ConvLSTM in parallel for the encoder model
  • CNN used for spatial information and ConvLSTM used for temporal information
  • CNN takes the input of last observed frame and ConvLSTM takes input of the difference between consecutive video frames

TODO:

  • Completion of encoder model
  • Work on decoder model

Week 5 - 6: Mar 1 to Mar 9, 2021

Minutes

  • Discussed the ways of implementing parallel processing in encoder
  • Discussed the possible ways to get the difference among video frames for the encoder input

TODO:

  • Try getting the difference among video frames using ffmpeg
  • Implement the parallel processing for encoder model

Week 4 - Feb 23, 2021

Minutes

  • Discussed in-depth functionality of Demo2Vec encoder and decoder

TODO:

  • Continue working on the Demo2Vec encoder

Week 3 - Feb 16, 2021

Minutes

  • Showed the improvement in Humanoid's hand through Mixamo animations

TODO:

  • Continue working on the Demo2Vec encoder
  • Discuss how the encoder and decoder is going to work in the model

Week 2 - Feb 9, 2021

Minutes

  • Showed the crooked hand of humanoid in dataset
  • Discussed plausible solutions and future plans

TODO:

  • Try to find a solution to fix the crooked hand of humanoid(Ethan)

Week 1 - Feb 2, 2021

Minutes

  • Showed and got CS298 - Proposal approved
  • Mentioned corrections required in the report and on the website

TODO:

  • Work on the comments mentioned for the proposal and get add-code

CS297

Week 16 - Dec 8, 2020

Minutes

  • Showed CS297-Report
  • Mentioned corrections required in the report and on the website

TODO:

  • Work on the comments mentioned for the report

Week 15 - Dec 1, 2020

Minutes

  • Worked towards and debugged OPRA dataset of Demo2Vec
  • Discussed ConvLSTM developed to handle video clips generated from OPRA dataset

TODO:

  • Work on CS297 Report

Week 14 - Nov 24, 2020

Minutes

  • Implemented Human Activity recognition on dataset UCF and HDMB to understand how to input video frames into model
  • Created two models, one model by extracting features from VVG16 and then passing them to convolutional layers and the other time series ConvLSTM model

TODO:

  • Work on implementing Demo2Vec model

Week 13 - Nov 17, 2020

Minutes

  • Showed movies of multiple scenes with multiple camera positions
  • Discussed various affordance papers and looked into Demo2Vec

TODO:

  • Try to implement simple neural network on video dataset similar to Demo2Vec

Week 12 - Nov 10, 2020

Minutes

  • Showed multiple scenes with multiple variations in the position of the 3D player(Ethan) and the position of pickable objects(like book, food etc.) in a room.
  • Showed the progress of Extended Deliverable 2
  • Showed video made out of multiple scenes and showed how the transitioning of scenes is happening with the script
  • Discussed how to make the whole dataset automated without the involvement of keys
  • Discussed the approach of using different camera positions to make 50 videos faster using script

TODO:

  • Continue working on Deliverable 2 (different camera positions)
  • Look into the researches going on in visual affordance

Week 11 - Nov 3, 2020

Minutes

  • Showed scene consisting of a 3D player(Ethan) and multiple objects(like book, food etc.) in a room.
  • Showed how player is able to walk towards the object and pick it using script
  • Recorded video of the whole scene
  • Deliverable 2 extended, create 50 videos of the scene using script

TODO:

  • Work on Extension of Deliverable 2

Week 10 - Oct 27, 2020

Minutes

  • Showed partial progress(placed objects and player to create an environment in Unity) of Deliverable 2
  • Discussed the plausible approach to reach the results of Deliverable 2

TODO:

  • Continuing working towards Deliverable 2

Week 9 - Oct 20, 2020

Minutes

  • Updated the progress on Unity
  • Discussed and resolved doubts regarding Deliverable 2

TODO:

  • Working towards Deliverable 2

Week 8 - Oct 13, 2020

Minutes

  • Showed the final model for object detection and classification of multi class in a single image using OpenCV and Keras
  • Deliverable 1 accepted

TODO:

  • Start working on Unity

Week 7 - Oct 6, 2020

Minutes

  • Showed model for object classification of multi class in a single image
  • Re-scoping the Deliverable to 2 phases: Emoji detection with Open CV and Emoji classification with Keras
  • Changes suggested in object classification model

TODO:

  • Divide the Deliverable into 2 phases
  • Incorporate the changes suggested

Week 6 - Sept 29, 2020

Minutes

  • Showed object detection and classification of 1 class in a single image
  • Presented Paper 1
  • Improvements suggested in Presentation of Paper 1

TODO:

  • Aim in making Object detection and classification multi-class

Week 5 - Sept 22, 2020

Minutes

  • Discussed possible approaches about Deliverable 1
  • Resolved issues regarding Deliverable 1
  • Paper suggested to study the technique used to detect objects without bounding boxes annotations(Paper 2)

TODO:

  • Progress on Deliverable 1
  • Presentation of Paper 1

Week 4 - Sept 15, 2020

Minutes

  • Showed object detection using OpenCV
  • Paper suggested to study the technique used to detect objects with OpenCV (Paper 1)

TODO:

  • Perform image classification using Pytorch
  • Complete Deliverable 1

Week 3 - Sept 8, 2020

Minutes

  • Updated the level of learning
  • Discussed changes and modifications in Deliverable 1

TODO:

  • Continue with learning OpenCV and Pytorch
  • Work towards deliverable#1

Week 2 - Sept 1, 2020

Minutes

  • Finalized CS297 topic
  • Discussed upcoming deliverables and possible scope of project

TODO:

  • Update proposal
  • Work towards deliverable#1

Week 1 - Aug 25, 2020

Minutes

  • First meeting to discuss possible topics
  • Topics discussed:-
    - Detect and predict affordance in a room
    - Mine Wiki data
    - Mine pattern among different mutations of disease using association rules
    - Compute K-means of neighboring data nodes in distributed network

TODO:

  • Write CS297 proposal