naomilwx / cs5242-project

Jupyter Notebooks

The notebooks documenting the data collection, data processing and models are in the notebook folder.

1. Data Collection (Naomi)
2. Data Exploration and Processing (Yi Xin)
   1. Data Exploration (Yi Xin)
   2. Data Preprocessing
      1. Background Removal (Naomi)
      2. Data selection (all)
3. Baseline CNN and Improvements (Naomi)
   1. Baseline model
   2. Improvements to baseline model
   3. Residual models
4. Exploration of Other Networks (Shubhankar)
   1. RNN+CNN model
   2. CNN+Attention models
      1. Local attention
      2. Convolution self attention

Data

The data used to train the model can be downloaded here

Dependencies

python: 3.9

Data analysis and processing

scikit-image: 0.19.3

scipy: 1.9.3

opencv-python: 4.6.0.66

torchvision; 0.13.1

torchtext: 0.13.1

imagehash: 4.2.1

matplotlib: 3.6.0

pandas: 1.4.3

rembg

Model training

numpy: 1.23.4

matplotlib: 3.6.0

opencv-python: 4.6.0.66

torch: >=1.12.1

torchvision; 0.13.1

tqdm

Project presentation slides

The project presentation slides are available here