General Information

[sab3rclaw1]

Hi, I am using Yolov3 model to train my dataset. I was wondering if you help me explain what does the output actually says. 1) How are bounding box location estimated? Is there any formula or code that I am missing or more precisely how is percentage_probability calculated? 2) What is the meaning of the number over bounding box in detected image? Is it IoU?

Please help

[OlafenwaMoses]

• The percentage probability of an object is the confidence score that the object is what it is classified to be. Objects with confidence less than the mininum_percentage_probability will be ignored
• The IoU is a value used during training/evaluation to determine the proximity of the generated bounding box values and the ground truth.
• The score displayed on a detected image is the object's class confidence score.

[dsdeeptisharma]

Hi, I just have a question. Will this ImageAI recognizes food images too? Can it tell that its a banana or orange after seeing the image? Thanks.

[OlafenwaMoses]

You have to

• collect images on the food classes you want to recognize (recommended 300 or above per class)
• train a model recognition/detection model using ImageAI

Recognition: https://github.com/OlafenwaMoses/ImageAI/blob/master/imageai/Prediction/CUSTOMTRAINING.md

Detection: https://github.com/OlafenwaMoses/ImageAI/blob/master/imageai/Detection/Custom/CUSTOMDETECTIONTRAINING.md

• apply the model on new images using ImageAI