I want to know the right Tensorflow object detection- api model for plant disease detection and weed detection?

Question

I have tried few zoo model faster_rcnn_resnet50_coco, but am not getting the desired output.

I am considering the below option Mask RCNN , but I don't know the right one to use, also am not sure I labelled my images correctly given the below output.

My labeling is focused more on detecting diseases on various part on the image rather that classifying the entire image as disease

mask_rcnn_resnet50_atrous_coco

mask_rcnn_inception_resnet_v2_atrous_coco

mask_rcnn_inception_v2_coco

mask_rcnn_resnet101_atrous_coco

enter image description here

Answer 1

From your output, annotations seem to be for object detection.

• For image segmentation you can prepare your data using VGG Image Annotator.
• You can try with U-Net model for image segmentation if models in models zoo are not giving desired output.

The basic difference between Object detection and Image segemntation is as follows.

• Object Detection:

  Object Detection extends localization to the next level where now the image is not constrained to have only one object, but can contain multiple objects. The task is to classify and localize all the objects in the image. Here again the localization is done using the concept of bounding box.

• Semantic Segmentation:

  The goal of semantic image segmentation is to label each pixel of an image with a corresponding class of what is being represented. Because we’re predicting for every pixel in the image, this task is commonly referred to as dense prediction.

In your case, you can try segmentation approach using U-Net.