Hong Chen
Hong Chen
I can do inference with Tensorflow 1.13 + python 3.5.2 and get reasonable output with pre-trained model. However, when I trained a model with flying chair dataset, the result didn't...
In the performance graph, is the single frame detector shown as "Dense Detection"? If yes, why ST-Lattice can achieve much better mAP compared to single frame detector? For DFF, the...
Could you elaborate on how to get the numbers on table 4? Is demo.py / demo_batch.py the right way to evaluate? From table 4, the speedup between DFF and Frame...
It seems to me the rescale_grad should be set to 1/batch_size. In case of 4 GPUs training rescale_grad should be 0.25. Why is the value set to "1.0" in train_end2end.py?
In the paper, the learning rate is 0.001. However, in the source code, the learning rate is 0.00025, as set in the yaml file. Why there is such a difference?
Hi, I'm reading the code to calculate mAP. In the code: https://github.com/msracver/Flow-Guided-Feature-Aggregation/blob/master/lib/dataset/imagenet_vid_eval.py#L58 When calculating the AUC of precision over recall on one class, it seems like the code enforces the...