Novel CNN Architecture for Semantic Segmentation of Autonomous Driving Videos
Published in Under submission, 2023
This work began as my undergraduate thesis at BUET, supervised by Dr. Md. Monirul Islam. It introduces two pooling methods for CNN-based semantic segmentation:
- Wavelet pooling preserves fine features and edges.
- Pixel binning reduces feature-map resolution to widen the receptive field.
Both methods are evaluated on the CamVid benchmark.
Code: GitHub
Recommended citation: Nahian Salsabil, Md. Monirul Islam. "Novel CNN Architecture for Semantic Segmentation of Autonomous Driving Videos." Under submission.
