自定义数据集上的tensorflow对象检测API评估

Question

I have a custom dataset that I trained using tensorflow object-detection api [Faster RCNN].

I saved the model checkpoints. Now, I want to run this model on my eval set. But, I cannot find the script that I use to run on eval set. I have tfrecord files for eval set.

How do I run the eval script from the object-detection api for my own saved model checkpoint? I couldn't find the script.

Answer 1

You are looking fot the eval.py .

Run this script with the command:

python eval.py \
--logtostderr \
--pipeline_config_path=PATH_TO_YOUR_CONFIG_FILE \
--checkpoint_dir=PATH_TO_CHECKPOINT \
--eval_dir=OUTPUT_PATH