SPECTR: Segmentation using Positional Encoding, CNNs, and Transformers

SPECTR by Anshul Ahluwalia, Aditya Singh, and Ranveer Thind

Visual Results (Cityscapes Image and Predicted Output):

To install SPECTR, we first much create a conda environment with the appropriate libraries.

conda env create -f environment.yml
conda activate spectr-v1

Next, install spectr as a module:

pip install -e .

Download the Cityscapes Dataset onto your local machine and unzip the files: leftImg8bit_trainvaltest.zip and gtFine_trainvaltest.zip.
To preprocess the image data and labels, run python spectr/src/cityscapes_preprocess.py. Before running this script, make sure to modify the file paths to be correct for your machine.
Modify the hyperparameters defined in spectr/config/config.yaml.
Finally, run python spectr/src/lightning_train.py to train the model using the Pytorch Lightning Framework. As training proceeds, a recent copy of your model will be saved in the lightning_logs/checkpoints directory.

To evaluate the SPECTR model checkpoint on a random training or testing batch, run python spectr/src/lightning_test.py.
- This will calculate mIoU for the model on the specific batch and dump all visualizations inside the viz folder.
- You must specify the absolute file path to the pytorch checkpoint model.