Skip to content

Review#6

Open
sahnilab wants to merge 8 commits intoitsasimiqbal:masterfrom
sahnilab:master
Open

Review#6
sahnilab wants to merge 8 commits intoitsasimiqbal:masterfrom
sahnilab:master

Conversation

@sahnilab
Copy link
Copy Markdown

No description provided.

Comment thread model.py
# indices = tf.image.non_max_suppression(
# normalized_boxes, scores, self.proposal_count,
# self.nms_threshold, name="rpn_non_max_suppression")
indices = tf.image.non_max_suppression(
Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This file contains original SeBRe model logic, and results in invalid detection results, even when running detection using weights that were also trained with normal non_max_suppression.

invalid

Comment thread SeBRe_detection.ipynb
"POOL_SIZE 7\n",
"POST_NMS_ROIS_INFERENCE 1000\n",
"POST_NMS_ROIS_TRAINING 2000\n",
"ROI_POSITIVE_RATIO 0.33\n",
"RPN_ANCHOR_RATIOS [0.5, 1, 2]\n",
"RPN_ANCHOR_SCALES (32, 64, 128, 256, 512)\n",
"RPN_ANCHOR_SCALES (16, 32, 64, 128, 256)\n",
Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I played with a few different factors of these values, but they did not seem to improve the results.

Comment thread model_softnms.py

# Non-max suppression
def nms(normalized_boxes, scores):
indices, scores = tf.image.non_max_suppression_with_scores(
Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This matches the call I made when generating the initial "softmax" weights, though running detection with the corresponding weights does not yield any results.

@sahnilab sahnilab marked this pull request as draft May 12, 2023 05:10
@sahnilab sahnilab marked this pull request as ready for review May 12, 2023 05:11
@sahnilab sahnilab marked this pull request as draft May 12, 2023 05:13
@sahnilab sahnilab marked this pull request as ready for review May 12, 2023 05:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant