This project implements a UNET-based semantic segmentation model for road scene segmentation. We utilized knowledge distillation to enhance the training process by transferring knowledge from a larger teacher model (Resnet34 backbone) to a smaller student model (Resnet18 backbone). The objective was to maintain high accuracy while reducing the model size for efficient deployment.
Teacher Model: UNET with Resnet34 backbone Student Model: UNET with Resnet18 backbone Knowledge Distillation: Online distillation, where both models are trained simultaneously and the student model learns from the teacher’s feature maps and predictions.
This dataset provides data images and labeled semantic segmentations captured via CARLA self-driving car simulator. The data was generated as part of the Lyft Udacity Challenge . This dataset can be used to train ML algorithms to identify semantic segmentation of cars, roads etc in an image.
We applied online knowledge distillation, where both the teacher and student models are trained in parallel. The student model receives two types of guidance:
Distillation Loss: Leveraging the softer output of the teacher model to guide the student, helping it learn intermediate features that improve its generalization.
Train Loss: 0.157
Train IOU: 0.650
Validation Loss: 0.210
Validation IOU: 0.623
The model achieves a good balance between accuracy and computational efficiency, with the student model reaching a Train IOU of 0.650 and a Validation IOU of 0.623.
Future Improvements: Fine-tuning hyperparameters to improve generalization. Experimenting with other teacher-student combinations or backbone architectures. Exploring post-processing techniques to enhance segmentation accuracy.
Logging the Experiments in WANDB
