Trying to upload multiple TRAINING images with custom vision SDK
for i in range(0,len(split_input)):
img_byte=connection_blob(confidential) #first I download images from my blob with a custom function, images names are in list split_input
image_list.append(ImageFileCreateEntry(name=split_input[i], contents=img_byte, tag_ids=[hemlock_tag])) #Then I create the list of images
for i in range(0,len(image_list)):
upload_result = trainer.create_images_from_files(model_id, images=[image_list[i]],raw=True) #I push data on custom vision
My problem is sometimes it goes wrong and only some images are send. For exemple I'm trying to send 20 pictures and only 10 are sended...
Any idea of where to look to try to fix the problem?
Edit: Even when I try to send images 1 by 1, they are not uploaded
I might have found the solution. Image are taken from a powerApps, and to make my test I was taking picture very fast. Some picture where strictly identical = custom vision only accept one version of the images.
You can check upload status with:
for i in range(0,len(image_list)):
upload_result = trainer.create_images_from_files(model_id, images=[image_list[i]],raw=True)
res_upload_img=upload_result.output.images
id_list.append(res_upload_img[0].image.id) #Get ID of uploaded image so u can track them on custom vision
img_status.append(res_upload_img[0].status) #return status i.e for example "Ok" or "OkDuplicated" if images was already existing