Hi! Thank you so much! Yes, you are right: ViT was only the beginning. It'd be very interesting to touch works such as the Swin transformer or DeiT. I'll keep in mind for future videos.
Yes, absolutely. I don't know how better than other more common solutions. You can start reading the section on segmentation of the original paper: arxiv.org/pdf/2104.14294.pdf
@@Mai-he2hv Absolutely! You should simply load weights from the torch hub and fine-tune them with a training loop. You can refer to "vision_transformers.py" to load weights and create the model.
Yes, you can do instance segmentation with transformers. Check this CVPR paper out: openaccess.thecvf.com/content/CVPR2021/papers/Wang_End-to-End_Video_Instance_Segmentation_With_Transformers_CVPR_2021_paper.pdf
Hi Altaf! Yes, it's in the video description. Anyway, this's the link of the notebook: github.com/EscVM/EscVM_YT/blob/master/Notebooks/2%20-%20PT1.X%20DeepAI-Quickie/pt_1_vit_attention.ipynb
finally someone explaining key,values,query notation in a simple and clear way. god may bless you.
thank you so much for this video, you also plan to make an explained video for the SWIN transformer? and relate it to these wonderfoul vids?
Hi! Thank you so much! Yes, you are right: ViT was only the beginning. It'd be very interesting to touch works such as the Swin transformer or DeiT. I'll keep in mind for future videos.
Wow Thanks for sharing !!
Great video! Waiting for more :)
Indeed true.
Great!!
Great video
is dino practical to use for segmentation of overlapping objects
Yes, absolutely. I don't know how better than other more common solutions. You can start reading the section on segmentation of the original paper: arxiv.org/pdf/2104.14294.pdf
@@escvm is there a way to fine-tune dino. also what do you recommend doing to use the model for instance segmentation
@@Mai-he2hv Absolutely! You should simply load weights from the torch hub and fine-tune them with a training loop. You can refer to "vision_transformers.py" to load weights and create the model.
Yes, you can do instance segmentation with transformers. Check this CVPR paper out: openaccess.thecvf.com/content/CVPR2021/papers/Wang_End-to-End_Video_Instance_Segmentation_With_Transformers_CVPR_2021_paper.pdf
Sir there is no link in the discription?
Hi Altaf! Yes, it's in the video description. Anyway, this's the link of the notebook: github.com/EscVM/EscVM_YT/blob/master/Notebooks/2%20-%20PT1.X%20DeepAI-Quickie/pt_1_vit_attention.ipynb