Unable to evaluate mAP scores

[willm9]

Hi, I've just been following along the documentation here: https://imageai.readthedocs.io/en/latest/customdetection/ to try and train a custom object detector.

I've just finished training and now want to evaluate the models. However, I'm just getting outputs like this:

---

Evaluation samples: 97 Using IoU: 0.5 Using Object Threshold: 0.3 Using Non-Maximum Suppression: 0.5 QIS: 0.0000 mAP: 0.0000

WARNING:tensorflow:No training configuration found in the save file, so the model was not compiled. Compile it manually.

---

I'm not sure where to go from here and hope I don't have to retrain everything again.

[ekesdf]

Did you set the JSON path correctly? The JSON file is in the cache folder created by image ai itself

[willm9]

Yes, I've set the JSON path correctly and can see it in the cache folder

[TheBeastCoding]

Here is a potential workaround if you are getting an mAP of 0.0 at the evaluation step (and have verified the h5 model and the training/evaluation data is not corrupted):

• Flush the workspace cache with: !rm -r .../<name_of_workspace>/cache/ During the evaluation step, ImageAI will recreate this folder along with the .pkl files. Between model training and evaluation, sometimes the file becomes "corrupted", which will give a default mAP of 0.0. Flushing the cache after training resolved this issue every time it occurred for me.