Add ONNX file download links to Per-Model Conversion Details table

README.md

@@ -114,15 +114,15 @@ All models verified against PyTorch source (absolute difference < 0.01):
 
 ### Per-Model Conversion Details
 
-| Model | Paper | Conference | Export Notes |
-|-------|-------|------------|-------------|
-| [LIQE](https://arxiv.org/abs/2304.00451) | *Blind Image Quality Assessment via Vision-Language Correspondence* | CVPR 2023 | Two-stage: CLIP ViT-B/32 image encoder + LIQE scoring head; text features pre-encoded to JSON |
-| [DBCNN](https://arxiv.org/abs/1907.02665) | *Blind Image Quality Assessment Using A Deep Bilinear CNN* | IEEE TCSVT 2020 | VGG16 + SCNN with bilinear pooling; both sub-networks exported as single model |
-| [HyperIQA](https://arxiv.org/abs/2003.08932) | *Blindly Assess Image Quality in the Wild Boosted by A Large-scale Database* | CVPR 2020 | ResNet50 backbone with hyper-network; exported forward patch path only |
-| [MANIQA](https://arxiv.org/abs/2204.08958) | *MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment* | CVPR 2022 NTIRE | ViT-B/8 backbone; single-file export (no external data) |
-| [MUSIQ](https://arxiv.org/abs/2108.05997) | *MUSIQ: Multi-scale Image Quality Transformer* | ICCV 2021 | Simplified to single-scale 224×224 input (original uses multi-scale patches) |
-| [TReS](https://arxiv.org/abs/2108.06858) | *No-Reference Image Quality Assessment via Transformers, Relative Ranking, and Self-Consistency* | WACV 2022 | Eval-only path; dual-path consistency and flip branches removed |
-| [CLIPIQA](https://arxiv.org/abs/2207.12396) | *Exploring CLIP for Assessing the Look and Feel of Images* | AAAI 2023 | CLIP-IQA+ variant with learned prompts baked into the model; antialias RN50 backbone |
+| Model | Paper | Conference | ONNX Files | Export Notes |
+|-------|-------|------------|------------|-------------|
+| [LIQE](https://arxiv.org/abs/2304.00451) | *Blind Image Quality Assessment via Vision-Language Correspondence* | CVPR 2023 | [`clip_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/clip_model.onnx) [`liqe_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/liqe_model.onnx) [`text_features.json`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/text_features.json) | Two-stage: CLIP ViT-B/32 image encoder + LIQE scoring head; text features pre-encoded to JSON |
+| [DBCNN](https://arxiv.org/abs/1907.02665) | *Blind Image Quality Assessment Using A Deep Bilinear CNN* | IEEE TCSVT 2020 | [`dbcnn_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/dbcnn_model.onnx) | VGG16 + SCNN with bilinear pooling; both sub-networks exported as single model |
+| [HyperIQA](https://arxiv.org/abs/2003.08932) | *Blindly Assess Image Quality in the Wild Boosted by A Large-scale Database* | CVPR 2020 | [`hyperiqa_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/hyperiqa_model.onnx) | ResNet50 backbone with hyper-network; exported forward patch path only |
+| [MANIQA](https://arxiv.org/abs/2204.08958) | *MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment* | CVPR 2022 NTIRE | [`maniqa_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/maniqa_model.onnx) | ViT-B/8 backbone; single-file export (no external data) |
+| [MUSIQ](https://arxiv.org/abs/2108.05997) | *MUSIQ: Multi-scale Image Quality Transformer* | ICCV 2021 | [`musiq_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/musiq_model.onnx) | Simplified to single-scale 224×224 input (original uses multi-scale patches) |
+| [TReS](https://arxiv.org/abs/2108.06858) | *No-Reference Image Quality Assessment via Transformers, Relative Ranking, and Self-Consistency* | WACV 2022 | [`tres_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/tres_model.onnx) | Eval-only path; dual-path consistency and flip branches removed |
+| [CLIPIQA](https://arxiv.org/abs/2207.12396) | *Exploring CLIP for Assessing the Look and Feel of Images* | AAAI 2023 | [`clipiqa_model.onnx`](https://huggingface.co/86Cao/IQA-ONNX-Models/blob/main/clipiqa_model.onnx) | CLIP-IQA+ variant with learned prompts baked into the model; antialias RN50 backbone |
 
 ## File List