Open tungnat97 opened 3 years ago
Hi this might be a bit unrelated but could you provide any insight on the config you used while training the centernet mobilenetv2. I am struggling on this for days, the training loss seems to decrease and everything seems fine while training but the model fails to give predicitions on val set and the mAP is close to zero.
Prerequisites
Please answer the following questions for yourself before submitting an issue.
1. The entire URL of the file you are using
CenterNet MobileNet V2
2. Describe the bug
When running following command:
python3 models/research/object_detection/export_tflite_graph_tf2.py --pipeline_config_path centernet_mobilenetv2_fpn_kpts/pipeline.config --trained_checkpoint_dir centernet_mobilenetv2_fpn_kpts/checkpoint/ --output_directory centernet_mobilenetv2_fpn_kpts/tflite/ --centernet_include_keypoints true --keypoint_label_map_path centernet_mobilenetv2_fpn_kpts/label_map.txt --max_detections 1 --config_override "model { center_net { image_resizer { fixed_shape_resizer { height: 256 width: 256 } } } }"
, I get the following error message:3. Steps to reproduce
pip install tf-nightly
.cd models/research/
protoc object_detection/protos/*.proto --python_out=.
cp object_detection/packages/tf2/setup.py .
python3 -m pip install .
python3 models/research/object_detection/export_tflite_graph_tf2.py --pipeline_config_path centernet_mobilenetv2_fpn_kpts/pipeline.config --trained_checkpoint_dir centernet_mobilenetv2_fpn_kpts/checkpoint/ --output_directory centernet_mobilenetv2_fpn_kpts/tflite/ --centernet_include_keypoints true --keypoint_label_map_path centernet_mobilenetv2_fpn_kpts/label_map.txt --max_detections 1 --config_override "model { center_net { image_resizer { fixed_shape_resizer { height: 256 width: 256 } } } }"
4. Expected behavior
TFLite model converted properly.
5. Additional context
Include any logs that would be helpful to diagnose the problem.
6. System information