Some problems with saving weights about pytorch-yolov3 HOT 3 CLOSED

Yes, I will push a fix for this later today.

from pytorch-yolov3.

Using the weights I trained , I'll get :

Performing object detection:
+ Batch 0, Inference Time: 0:00:00.262377
+ Batch 1, Inference Time: 0:00:00.026853
+ Batch 2, Inference Time: 0:00:00.032170
+ Batch 3, Inference Time: 0:00:00.008161
+ Batch 4, Inference Time: 0:00:00.007893
+ Batch 5, Inference Time: 0:00:00.008063
+ Batch 6, Inference Time: 0:00:00.007226
+ Batch 7, Inference Time: 0:00:00.007327
+ Batch 8, Inference Time: 0:00:00.007915
+ Batch 9, Inference Time: 0:00:00.008136
+ Batch 10, Inference Time: 0:00:00.008143
+ Batch 11, Inference Time: 0:00:00.006397
+ Batch 12, Inference Time: 0:00:00.006317
+ Batch 13, Inference Time: 0:00:00.005665
+ Batch 14, Inference Time: 0:00:00.005724

Does this means no prediction has been made? The loss is dropping during the training, I used a weight which has 0.3 total loss to predict. I wonder if there is an issue in the loss function or somewhere else.

from pytorch-yolov3.

Hi, this is fixed in 6727086. Unfortunately I have yet to get the model to converge to the same results as the authors during training.

from pytorch-yolov3.