Card

CluelessCoders AutoWCEBleedGen Challenge MISAHUB

Welcome to CluelessCoders’ GitHub repository on Auto-WCEBleedGen Challenge MISAHUB 2023! This repository contains code, datasets, and a stored model for the image classification as bleeding or non-bleeding. We also have an Excel sheet containing the image IDs and predicted class labels for testing datasets 1 and 2.

Finalised ColabNotebook is MisaHubResNet_FINAL.ipynb

Achieved evaluation metrics of the validation dataset.

Classification -

Parameter Score
Accuracy 0.99
Recall 0.99
F1 - Score 0.99

Detection -

Parameter Score
Mean average precision 0.58
Intersection over union (IoU) (AVERAGE) 0.44
Average precision -

Screenshot of the best 5 predicted frames with bounding boxes and confidence level from DATASET 1 (including the CI in the screenshot itself)-

1_3

1_1

1_4

1_2

1_5

Screenshot of the best 5 predicted frames with bounding boxes and confidence level from DATASET 2 (including the CI in the screenshot itself)-

2_1
2_2
2_5
2_4
2_3

Stored Model drive link with access : https://drive.google.com/drive/folders/1s5poG5v3myJqoR2GBYZwbCbMeDJ9wSSu?usp=sharing
resnet_model_MISAHUB for classification

best.pt for yolov5