CODEWITHSUNDEEP
pythontensorflow
Irdina Hidayah
Irdina Hidayah

Reputation: 11

How to get the accuracy from the evaluation of a pretrained model using Tensorflow object detection api?

I'm using Tensorflow object detection API models for my plate number detection project. I'm using MobileNet SSD pre-trained model ('ssd_mobilenet_v2_fpnlite_320x320_coco17_tpu-8'). I have evaluated my pre-trained models and only get the average precision and recall. Is there any other way I can get the value of the accuracy model or confusion matrix?

This is the command code of the evaluation model

command = "python {} --model_dir={} --pipeline_config_path={} --checkpoint_dir={}".format(TRAINING_SCRIPT, paths['CHECKPOINT_PATH'],files['PIPELINE_CONFIG'], paths['CHECKPOINT_PATH'])

This is the output

 Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.543641
 Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 1.00000
 Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.623451
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.570156
 Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.566508
 Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.514109
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.573684
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.636842
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.636842
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.666667
 Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.630769
 Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.633333

Upvotes: 0

Views: 344

Answers (1)

user11530462
user11530462

Reputation:

  • To get the confusion matrix for the object detection model you have to find the Intersection over union(IoU) for the predictions.

  • IoU is defined as the area of intersection between ground truth mask and predicted mask divided by the area of the union between the two.

  • Based on the calculated IoU you have to define a threshold to get the True positive, False positive, True negative, False negative.

    For more details please refer to this link. Thank You!

Upvotes: 1

Related Questions