Available scripts:

• compute_computational_complexity.py: computes the GFlops per model and stores the results into a JSON file;
• compute_memory_usage.py: estimates the GPU-memory usage per model by varying the batch size (requires the installation of GPUtil module). It stores the results into a JSON file;
• compute_inference_time.py: computes inference time per model by varying the batch size (resulting inference time is the average of 10 runs);
• compute_accuracy_rate.py: estimates performances on Imagenet validation set per model and stores the results into a JSON file.