Scores and boxes of onnx model exporting by jetson-inference

Hi,everyone! I am confused about jetson-inference project. I had re-trained ssd-mobilenet model, and converted it to onnx.And scores, the output of onnx model that exporting by runing onnx_export.py, seems different from confidence what we used to measure model’s performance. I even found that some scores are minus! Why?

Hi,

Could you help to confirm if the training process is succeed or not first.
You should get some accuracy value when applying the training like below:

2020-07-10 13:14:12 - Epoch: 0, Step: 10/1287, Avg Loss: 12.4240, Avg Regression Loss 3.5747, Avg Classification Loss: 8.8493
2020-07-10 13:14:12 - Epoch: 0, Step: 20/1287, Avg Loss: 9.6947, Avg Regression Loss 4.1911, Avg Classification Loss: 5.5036
2020-07-10 13:14:13 - Epoch: 0, Step: 30/1287, Avg Loss: 8.7409, Avg Regression Loss 3.4078, Avg Classification Loss: 5.3332
2020-07-10 13:14:13 - Epoch: 0, Step: 40/1287, Avg Loss: 7.3736, Avg Regression Loss 2.5356, Avg Classification Loss: 4.8379
2020-07-10 13:14:14 - Epoch: 0, Step: 50/1287, Avg Loss: 6.3461, Avg Regression Loss 2.2286, Avg Classification Loss: 4.1175
...
2020-07-10 13:19:26 - Epoch: 0, Validation Loss: 5.6730, Validation Regression Loss 1.7096, Validation Classification Loss: 3.9634
2020-07-10 13:19:26 - Saved model models/fruit/mb1-ssd-Epoch-0-Loss-5.672993580500285.pth

Thanks.

Thank you but I had sloved the problem. I made some mistake in my c++ code.

Okay. Thanks for the update.