computing precision and recall

[Nocoffeeanymore]

❔Question

Why does author set a pr_score=0.1 and utilize "np.interp(-pr_score, -conf[i], precision[:, 0])" to compute precision as well as recall?
Could anyone give me some help? Thanks.

Additional context

[github-actions]

Hello @Nocoffeeanymore, thank you for your interest in our work! Please visit our Custom Training Tutorial to get started, and see our Jupyter Notebook , Docker Image, and Google Cloud Quickstart Guide for example environments.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@Nocoffeeanymore you can set it to any value you want.

[Nocoffeeanymore]

Thank you very much.
# Sort by objectness
i = np.argsort(-conf)
tp, conf, pred_cls = tp[i], conf[i], pred_cls[i]
r[ci] = np.interp(-pr_score, -conf[i], recall[:, 0]) # r at pr_score, negative x, xp because xp decreases
Maybe the second conf[i] is not monotonically decreasing. Then it becomes difficult to understand.

[glenn-jocher]

@Nocoffeeanymore conf is already sorted here.

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[diego0718]

Hi all,

How can i modify the train script to compute the precision vs recall curve and graph it ?

[glenn-jocher]

@diego0718 that's a good question. This should really be a training output.

All the required plotting code is here, you simply uncomment this to produce a PR curve:

yolov5/utils/general.py

Lines 295 to 304 in 9eae82e

	# Plot
	# fig, ax = plt.subplots(1, 1, figsize=(5, 5))
	# ax.plot(recall, precision)
	# ax.set_xlabel('Recall')
	# ax.set_ylabel('Precision')
	# ax.set_xlim(0, 1.01)
	# ax.set_ylim(0, 1.01)
	# fig.tight_layout()
	# fig.savefig('PR_curve.png', dpi=300)

TODO: autogenerate PR curve on final test.py run.

[glenn-jocher]

@diego0718 PR has been submitted for this feature addition, please review and comment at #1107

[diego0718]

@diego0718 that's a good question. This should really be a training output.

All the required plotting code is here, you simply uncomment this to produce a PR curve:

yolov5/utils/general.py

Lines 295 to 304 in 9eae82e

	# Plot
	# fig, ax = plt.subplots(1, 1, figsize=(5, 5))
	# ax.plot(recall, precision)
	# ax.set_xlabel('Recall')
	# ax.set_ylabel('Precision')
	# ax.set_xlim(0, 1.01)
	# ax.set_ylim(0, 1.01)
	# fig.tight_layout()
	# fig.savefig('PR_curve.png', dpi=300)

TODO: autogenerate PR curve on final test.py run.

Thank u Glen for rapid response. Good work!