Closed prasanna-in closed 6 years ago
Looks Like PNG Processing Error
Yes, it looks like that PNG has only 1 channel (grayscale) instead of 3 (rgb), I'll try to fix it, meanwhile you can convert that png to rgb with some image tools
Yeah converting that image to JPG and removing alpha channel it works well, we suppose to handle that case, but that PNG returns a strange shape
Is fixed now docker pull machinebox/classificationbox:latest
to get the changes
classificationbox v0 (development)
http://localhost:8080/classificationbox/models/sentiment2/teach
{ "class": "Img_URL", "inputs": [ { "key": "user_age", "type": "image_url", "value": "https://db.creditcard/wp-content/uploads/2016/10/bankamericard_travel_rewards_credit_card.png"
}
Response :
{ "success": false, "error": "sentiment2:sentimentModel2 could not preprocess the input: could not pre process the image for feature 'user_age': model predict image returns an error Traceback (most recent call last):\n File \"/tmp/mb439721528\", line 49, in predict_img\n File \"/env/local/lib/python2.7/site-packages/keras/engine/training.py\", line 1695, in predict\n check_batch_axis=False)\n File \"/env/local/lib/python2.7/site-packages/keras/engine/training.py\", line 144, in _standardize_input_data\n str(array.shape))\nValueError: Error when checking : expected input_1 to have shape (None, None, None, 3) but got array with shape (1, 299, 299, 1)\n" }