Computer Vision Project

These are some mini projects I created using Pytorch for computer vision tasks

in this repo I will create a mini project with various tasks in computer vision such as image classification, object detection, object segmentation, GAN and many more

add a star to get updates from this repo

Repo Update

• 26 Sep 2024 -> inference next js plant-species-classification
• 25 Sep 2024 -> inference next js cat-and-dogs-classification
• 20 Sep 2024 -> facemask-detection
• 17 Sep 2024 -> manga-text-detection
• 17 Sep 2024 -> plant-species-classification
• 16 Sep 2024 -> cat-and-dogs-classification

Dont have CUDA for Training ? Download Model Here

Model Name	Model URL
Cat and Dogs	click here to download
Plant Species	click here to download
Manga Text	click here to download
Facemask Detection	click here to download

cat-and-dogs-classification (Created : 16-09-2024)

in this project, I focused on fine-tune the resnet-50 model then exported it to onnx format.

which initially resnet-50 outputs approximately 1000 classes into just 2 classes according to the dataset that has been downloaded, namely cats and dogs

• Loss : 0.06402150790199812
• Accuracy : 98.46762234305487%
• Epochs : 5

plant-species-classification (Created : 17-09-2024)

in this project, I focused on fine-tune the google/vit-base-patch16-224 model then exported it to onnx format.

which initially google/vit-base-patch16-224 outputs approximately 1000 classes into just 47 classes according to the dataset that has been downloaded, namely house-plant-species

• Loss : 0.305279920695395
• Accuracy : 93.55418434246046%
• Epochs : 5

manga-text-detection (Created : 17-09-2024)

in this project, I focused on fine-tune the faster-RCNN model then exported it to onnx format.

create object detection on text in manga pages with the manga-text-detection dataset

• Loss : 0.105800
• Epochs : 25
• MAP : 0.9
• MAP-50 : 1.0
• MAP-75 : 0.9901

facemask-detection (Created : 20-09-2024)

in this project, I focused on fine-tune the YOLOv10 model then exported it to onnx format.

create object detection on facemask-dataset

• Loss : 0.405800
• Epochs : 50
• MAP : 0.908
• MAP-50 : 0.934
• MAP-75 : 0.666

Todo List

• Fine tune resnet-50 on the cats and dogs dataset
  • Create inference with opencv and onnxruntime (python)
  • Create inference with next.js and onnxruntime-web
  • Create Inference with kotlin
• Fine tune vit-base-224 on the plant species dataset
  • Create inference with opencv and onnxruntime (python)
  • Create inference with next.js and onnxruntime-web
  • Create Inference with kotlin
• Fine tune Faster-RCNN on the manga text detection dataset
  • Create inference with opencv and onnxruntime (python)
  • Create inference with next.js and onnxruntime-web
• Fine tune YOLO-V10 on the facemask dataset
  • Create inference with opencv and onnxruntime (python)
  • Create inference with next.js and onnxruntime-web
  • Create Inference with kotlin and run with input uint8

Authors

• @naufalahnaf17