How to test model on a single image

Hi, How can I test model on new image? what kind of preprocessing I need to pass the image to the network? Is this correct?

image = cv2.imread('test/1.jpg', 1)

draw = image.copy()
draw = cv2.cvtColor(draw, cv2.COLOR_BGR2RGB)
image, scale = resize_image(image)
image = image.astype('float32')
image[:, :, 0] -= 103.939
image[:, :, 1] -= 116.779
image[:, :, 2] -= 123.68
image = image/np.max(image)
start = time.time()
_, _, detections = model.predict_on_batch(np.expand_dims(image, axis=0))
print detections
print("processing time: ", time.time() - start)
predicted_labels = np.argmax(detections[0, :, 4:], axis=1)
scores = detections[0, np.arange(detections.shape[1]), 4 + predicted_labels]
detections[0, :, :4] /= scale

in This case all of my predicted labels are zero. I don’t know it is for bad training or wrong test code.

[hgaiser]

The only problem is that you don’t need to rescale the values to [0, 1] (in other words, no need for image = image/np.max(image)). Other than that it looks fine. I’ll see if I can add an example like this to the jupyter notebook, since it is a more frequent question.

Can you let me know if that solves your issue ?

[hgaiser]

That’s hard to say, it depends on how difficult your dataset is really. But those loss values seem very reasonable to me.

I’ll close this since the original issue was resolved. For further discussions, please join the Slack channel.