Introduction
NeuralMarker offers a variety of state-of-the-art Pretrained eep Learning Models for Image Segmentation from different available frameworks like TensorFlow and PyTorch.
These available Pretrained deep learning models cover broad categories of Image Segmentation tasks such as :
- Semantic segmentation
- Instance segmentation
- Panoptic segmentation
This list of models at hand is not static, as more models in the Neural Network Library of NeuralMarker will be added, then accordingly the below list will be updated too:
Pretrained Models
1. DeepLabv3_MobileNetV2
Model Name: DeepLabv3
Network Backbone: MobileNetv2
No of Categories: 21
Convenient to Use: For Semantic Image Segmentation
Dataset Used: Pascal VOC (2007-2012)
2. DeepLabv3_xception65
Model Name: DeepLabv3
Network Backbone: xception65
No of Categories: 21
Convenient to Use: For Semantic Image Segmentation
Dataset Used: Pascal VOC (2007-2012)
3. Matterport_Mask R-CNN_MS COCO
Framework Used: Mask R-CNN
Network Backbone: ResNet101
No of Categories: 80
Convenient to Use: For Instance Image Segmentation
Dataset Used: MS COCO
4. DETR [Detection Transformer]
Framework Used: Faster R-CNN
Network Backbone: ResNet-50
No of Categories: 80
Convenient to Use: For Instance Image Segmentation, Panoptic Segmentation
Dataset Used: MS COCO
5. Detectron2_Cityscapes_MaskRCNN
Framework Used: Mask R-CNN
Network Backbone: ResNet-50 + Feature Pyramid Network (FPN)
No of Categories: 8
Convenient to Use: For Instance Image Segmentation
Dataset Used: Cityscapes
6. Detectron2_MS COCO_MaskRCNN
Framework Used: Mask R-CNN
Network Backbone: ResNet-50 + Feature Pyramid Network (FPN)
No of Categories: 80
Convenient to Use: For Instance Image Segmentation
Dataset Used: MS COCO
7. LVIS_R50FPN
Framework Used: Mask R-CNN
Network Backbone: ResNet-50 + Feature Pyramid Network (FPN)
No of Categories: 1230
Convenient to Use: For Vocabulary Instance Image Segmentation
Dataset Used: LVIS
8. LVIS_R101FPN
Framework Used: Mask R-CNN
Network Backbone: ResNet-101 + Feature Pyramid Network (FPN)
No of Categories: 1230
Convenient to Use: For Vocabulary Instance Image Segmentation
Dataset Used: LVIS
9. MMSEG_PSPNet_cityscape_r101
Framework Used: PSPNet
Network Backbone:R-101-D8
Dataset Used: Cityscapes
lr schld: 80000
Convenient to Use: Image Segmentation
10. MMSEG_PSPNet_cityscape_r50
Framework Used: PSPNet
Network Backbone:R-50-D8
Dataset Used: Cityscapes
lr schld: 80000
Convenient to Use: Image Segmentation
11. HRNet-Semantic-Segmentation-OCR_Cityscape
Network Backbone: HRNet-W48 + OCR
No of Categories: 19
Dataset Used: Cityscapes
Convenient to Use: Semantic Segmentation
12. HRNet-Semantic-Segmentation-OCR_pascal_ctx
Network Backbone: HRNet-W48 + OCR
No of Categories: 59
Dataset Used: PASCAL-Context
Convenient to Use: Semantic Segmentation
13. HRNet-Semantic-Segmentation-OCR_ade20k
Network Backbone: HRNet-W48 + OCR
No of Categories: 150
Dataset Used: ADE20K
Convenient to Use: Semantic Segmentation
14. HRNet-Semantic-Segmentation-OCR_COCOStuff
Network Backbone: HRNet-W48 + OCR
No of Categories: 171
Dataset Used: COCO-Stuff
Convenient to Use: Semantic Segmentation
15. HRNet-Semantic-Segmentation-OCR_LIP
Network Backbone: HRNet-W48 + OCR
No of Categories: 20 classes (19 semantic human part classes and 1 background class)
Dataset Used: LIP
Convenient to Use: Semantic Segmentation
16. HRNet-Semantic-Segmentation_Cityscape_small_v1
Model Name: HRNetV2-W18-Small-v1
No of Categories: 19
Dataset Used: Cityscapes
Convenient to Use: Semantic Segmentation
17. HRNet-Semantic-Segmentation_Cityscape_Big
Model Name: HRNetV2-W48
No of Categories: 19
Dataset Used: Cityscapes
Convenient to Use: Semantic Segmentation
18. HRNet-Semantic-Segmentation_LIP
Model Name: HRNetV2-W48
No of Categories: 20
Dataset Used: LIP
Convenient to Use: Semantic Segmentation
19. HRNet-Semantic-Segmentation_pascal_ctx
Model Name: HRNetV2-W48
No of Categories: 59
Dataset Used: PASCAL-Context
Convenient to Use: Semantic Segmentation
20. MMSEG_OCRNet_VOC12_HRNet48
Method Used: OCRNet
Network Backbone:HRNetV2p-W48
Dataset Used: PASCAL VOC 2012
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
21. MMSEG_OCRNet_VOC12_HRNet18_small
Method Used: OCRNet
Network Backbone:HRNetV2p-W18-small
Dataset Used: PASCAL VOC 2012
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
22. MMSEG_OCRNet_VOC12_HRNet18
Method Used: OCRNet
Network Backbone:HRNetV2p-W18
Dataset Used: PASCAL VOC 2012
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
23. MMSEG_OCRNet_ade20k_HRNet18_small
Method Used: OCRNet
Network Backbone:HRNetV2p-W18
Dataset Used: ADE20k
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
24. MMSEG_OCRNet_ade20k_HRNet18
Method Used: OCRNet
Network Backbone:HRNetV2p-W18
Dataset Used: ADE20k
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
25. MMSEG_OCRNet_ade20k_HRNet48
Method Used: OCRNet
Network Backbone:HRNetV2p-W48
Dataset Used: ADE20k
Input Resolution : 512x512
Convenient to Use: Semantic Segmentation
26. MMSEG_deeplabv3_pascal_ctx_r101
Method Used: DeepLabV3
Network Backbone: R-101-D8
No of Categories: 60
Dataset Used: PASCAL-Context
Convenient to Use: Semantic Segmentation
27. MMSEG_deeplabv3_VOC12_r101
Method Used: DeepLabV3
Network Backbone: R-101-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
28. MMSEG_deeplabv3_VOC12_r50
Method Used: DeepLabV3
Network Backbone: R-50-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
29. MMSEG_deeplabv3_ade20k_r101
Method Used: DeepLabV3
Network Backbone: R-101-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
30. MMSEG_deeplabv3_ade20k_r50
Method Used: DeepLabV3
Network Backbone: R-50-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
31. MMSEG_deeplabv3plus_pascal_ctx_r101
Method Used: DeepLabV3+
Network Backbone: R-101-D8
No of Categories: 60
Dataset Used: PASCAL-Context
Convenient to Use: Semantic Segmentation
32. MMSEG_deeplabv3plus_VOC12_r101
Method Used: DeepLabV3+
Network Backbone: R-101-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
33. MMSEG_deeplabv3plus_VOC12_r50
Method Used: DeepLabV3+
Network Backbone: R-50-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
34. MMSEG_deeplabv3plus_ade20k_r101
Method Used: DeepLabV3+
Network Backbone: R-101-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
35. MMSEG_deeplabv3plus_ade20k_r50
Method Used: DeepLabV3+
Network Backbone: R-50-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
36. MMSEG_PSPNet_pascal_ctx_r101
Method Used: PSPNet
Network Backbone: R-101-D8
No of Categories: 60
Dataset Used: PASCAL-Context
Convenient to Use: Semantic Segmentation
37. MMSEG_PSPNet_VOC12_r50
Method Used: PSPNet
Network Backbone: R-50-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
38. MMSEG_PSPNet_VOC12_r101
Method Used: PSPNet
Network Backbone: R-101-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
39. MMSEG_PSPNet_ade20k_r50
Method Used: PSPNet
Network Backbone: R-50-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
40. MMSEG_PSPNet_ade20k_r101
Method Used: PSPNet
Network Backbone: R-101-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
41. MMSEG_ANN_ade20k_r101
Method Used: ANN
Network Backbone: R-101-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
42. MMSEG_ANN_ade20k_r50
Method Used: ANN
Network Backbone: R-50-D8
No of Categories: 150
Dataset Used: ADE20k
Convenient to Use: Semantic Segmentation
43. MMSEG_ANN_VOC12_r50
Method Used: ANN
Network Backbone: R-50-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
44. MMSEG_ANN_VOC12_r101
Method Used: ANN
Network Backbone: R-101-D8
No of Categories: 21
Dataset Used: PASCAL VOC 2012
Convenient to Use: Semantic Segmentation
How to create Annotations on an Image Segmentation Dataset using S-O-T-A Pre-Trained Models
Steps to Follow
1. log in to the tool
2. Click on the Add button.
3. The Add Dataset Form will appear.
4. Fill in all the fields such as dataset name, dataset description, category-type, and categories.
4 a. Select category-type as Segmentation
4 b. Choose Pretraining Models from the available list.
4 c. Categories according to the chosen model will appear.
5. Add data to NeuralMarker using the available options listed below:
- Google Drive Link or S3 link
- CSV File with Image URL's
- Drag and Drop
6. Click on Submit Button
7. A New Dataset will be created.
8. After New Dataset is created:
- The Dataset card will display the status "Pretrain Model running".
- The brain symbol on hover will display "AI labeling in Progress".
9. After Pretrain Model stops running over the entire dataset:
- The Dataset card will display the status "Pretrain Model Done & Ready For Annotation".
- The brain symbol on the card will display "AI Labeling Report".
10. After Pretrain Model stops running over the entire dataset, an AI labeling report will be generated:
- With a confidence score of annotations over the images in the dataset.
