I did an experiment to compare how model performance change when converted to TensorRT engine. These are the steps I took.
1. Train Several Binary Segmentation Model
2. Randomly Choose 30 images from the dataset
3. Run inference with the chosen 30 images and calculate Average IoU
I’ve found out that the higher the capacity(How much a model can understand a complex task), the more likely the precision decreases dramatically when converted to engine files.
