trained model converted for inference/testing

[wiamadaya]

i converted both models the results from train.py and train_iou.py using convert_model.py, and use those models to perform inference/testing using this notebook from keras-retinanet https://github.com/delftrobotics/keras-retinanet/blob/master/examples/ResNet50RetinaNet.ipynb

Modified the imports accordingly from object_detector_retinanet.keras_retinanet and run the inference/testing

inference/testing using model converted from train.py went successfully, but inference/testing using model converted from train_iou.py i received the following error

---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-5-c86318e1b643> in <module>
12 # process image
13 start = time.time()
---> 14 boxes, scores, labels = model.predict_on_batch(np.expand_dims(image, axis=0))
15 print("processing time: ", time.time() - start)
16

ValueError: too many values to unpack (expected 3)

[yfe404]

This is because the model with iou outputs one more element, try this:

boxes, hard_scores, labels, soft_scores = model.predict_on_batch(np.expand_dims(image, axis=0))

[wiamadaya]

@yafeunteun thank you, i have made the modification on the for-loop as well, now it's working, but i'm still unsure whether the inference pipeline is correct, i mean how to properly utilize hard_scores and soft_scores on the prediction

[yfe404]

Here is a (quick and dirty) example that takes an image and plot bounding box detections as well as labels (tested with a single class):

https://github.com/yafeunteun/SKU110K_code/blob/master/examples/retinanet.ipynb
The first image is fed to retinanet, the second to the retinanet + iou merger aka the complete pipeline presented in the paper.

Here I've set the hard_score_rate argument to .3, from what I understand (I may be wrong, please double check), that means that the model will set the confidence that a box contains an object to .3hard_score + .7soft_score (both hard_score and soft_score are output by the model for each bounding box detected).

[wiamadaya]

@yafeunteun thanks for being so helpful

[Pari-singh]

The weights shared at #9 outputs only 3 elements. Wanted to confirm if hard and soft scores are merged in that?

[eg4000]

#30