Update README.md

README.md

@@ -8,9 +8,9 @@ license: mit
 <div align="center">
 
 [![Project Page](https://img.shields.io/badge/Project-Website-orange)](https://farlongctx.github.io/)
-[![arXiv](https://img.shields.io/badge/arXiv-2503.19325-b31b1b.svg)](https://arxiv.org/abs/2503.19325)&nbsp;
-[![huggingface weights](https://img.shields.io/badge/%F0%9F%A4%97%20Weights-FAR-yellow)](https://huggingface.co/guyuchao/FAR_Models)&nbsp;
-[![SOTA](https://img.shields.io/badge/State%20of%20the%20Art-Video%20Generation%20-32B1B4?logo=data%3Aimage%2Fsvg%2Bxml%3Bbase64%2CPHN2ZyB3aWR0aD0iNjA2IiBoZWlnaHQ9IjYwNiIgeG1sbnM9Imh0dHA6Ly93d3cudzMub3JnLzIwMDAvc3ZnIiB4bWxuczp4bGluaz0iaHR0cDovL3d3dy53My5vcmcvMTk5OS94bGluayIgb3ZlcmZsb3c9ImhpZGRlbiI%2BPGRlZnM%2BPGNsaXBQYXRoIGlkPSJjbGlwMCI%2BPHJlY3QgeD0iLTEiIHk9Ii0xIiB3aWR0aD0iNjA2IiBoZWlnaHQ9IjYwNiIvPjwvY2xpcFBhdGg%2BPC9kZWZzPjxnIGNsaXAtcGF0aD0idXJsKCNjbGlwMCkiIHRyYW5zbGF0ZSgxIDEpIj48cmVjdCB4PSI1MjkiIHk9IjY2IiB3aWR0aD0iNTYiIGhlaWdodD0iNDczIiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iMTkiIHk9IjY2IiB3aWR0aD0iNTciIGhlaWdodD0iNDczIiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iMjc0IiB5PSIxNTEiIHdpZHRoPSI1NyIgaGVpZ2h0PSIzMDIiIGZpbGw9IiM0NEYyRjYiLz48cmVjdCB4PSIxMDQiIHk9IjE1MSIgd2lkdGg9IjU3IiBoZWlnaHQ9IjMwMiIgZmlsbD0iIzQ0RjJGNiIvPjxyZWN0IHg9IjQ0NCIgeT0iMTUxIiB3aWR0aD0iNTciIGhlaWdodD0iMzAyIiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iMzU5IiB5PSIxNzAiIHdpZHRoPSI1NiIgaGVpZ2h0PSIyNjQiIGZpbGw9IiM0NEYyRjYiLz48cmVjdCB4PSIxODgiIHk9IjE3MCIgd2lkdGg9IjU3IiBoZWlnaHQ9IjI2NCIgZmlsbD0iIzQ0RjJGNiIvPjxyZWN0IHg9Ijc2IiB5PSI2NiIgd2lkdGg9IjQ3IiBoZWlnaHQ9IjU3IiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iNDgyIiB5PSI2NiIgd2lkdGg9IjQ3IiBoZWlnaHQ9IjU3IiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iNzYiIHk9IjQ4MiIgd2lkdGg9IjQ3IiBoZWlnaHQ9IjU3IiBmaWxsPSIjNDRGMkY2Ii8%2BPHJlY3QgeD0iNDgyIiB5PSI0ODIiIHdpZHRoPSI0NyIgaGVpZ2h0PSI1NyIgZmlsbD0iIzQ0RjJGNiIvPjwvZz48L3N2Zz4%3D)](https://paperswithcode.com/sota/video-generation-on-ucf-101)
+[![arXiv](https://img.shields.io/badge/arXiv-2503.19325-b31b1b.svg)](https://arxiv.org/abs/2503.19325)
+[![huggingface weights](https://img.shields.io/badge/%F0%9F%A4%97%20Weights-FAR-yellow)](https://huggingface.co/guyuchao/FAR_Models)
+[![SOTA](https://img.shields.io/badge/State%20of%20the%20Art-Video%20Generation%20-32B1B4)](https://paperswithcode.com/sota/video-generation-on-ucf-101)
 
 </div>
 
@@ -18,7 +18,7 @@ license: mit
   <a href="https://arxiv.org/abs/2503.19325">Long-Context Autoregressive Video Modeling with Next-Frame Prediction</a>
 </p>
 
-![dmlab_sample](https://github.com/showlab/FAR/blob/main/assets/dmlab_sample.png)
+![dmlab_sample](https://github.com/showlab/FAR/blob/main/assets/dmlab_sample.png?raw=true)
 
 ## 📢 News
 
@@ -31,12 +31,12 @@ license: mit
 
 FAR (i.e., <u>**F**</u>rame <u>**A**</u>uto<u>**R**</u>egressive Model) learns to predict continuous frames based on an autoregressive context. Its objective aligns well with video modeling, similar to the next-token prediction in language modeling.
 
-![dmlab_sample](./assets/pipeline.png)
+![dmlab_sample](https://github.com/showlab/FAR/blob/main/assets/pipeline.png?raw=true)
 
 ### 🔥 FAR achieves better convergence than video diffusion models with the same continuous latent space
 
 <p align="center">
-<img src="./assets/converenge.jpg" width=55%>
+<img src="https://github.com/showlab/FAR/blob/main/assets/converenge.jpg?raw=true" width=55%>
 <p>
 
 ### 🔥 FAR leverages clean visual context without additional image-to-video fine-tuning:
@@ -44,19 +44,19 @@ FAR (i.e., <u>**F**</u>rame <u>**A**</u>uto<u>**R**</u>egressive Model) learns t
 Unconditional pretraining on UCF-101 achieves state-of-the-art results in both video generation (context frame = 0) and video prediction (context frame ≥ 1) within a single model.
 
 <p align="center">
-<img src="./assets/performance.png" width=75%>
+<img src="https://github.com/showlab/FAR/blob/main/assets/performance.png?raw=true" width=75%>
 <p>
 
 ### 🔥 FAR supports 16x longer temporal extrapolation at test time
 
 <p align="center">
-<img src="./assets/extrapolation.png" width=100%>
+<img src="https://github.com/showlab/FAR/blob/main/assets/extrapolation.png?raw=true" width=100%>
 <p>
 
 ### 🔥 FAR supports efficient training on long-video sequence with managable token lengths
 
 <p align="center">
-<img src="./assets/long_short_term_ctx.jpg" width=55%>
+<img src="https://github.com/showlab/FAR/blob/main/assets/long_short_term_ctx.jpg?raw=true" width=55%>
 <p>
 
 #### 📚 For more details, check out our [paper](https://arxiv.org/abs/2503.19325).