Document model attribution and UBC results

README.md

@@ -37,19 +37,42 @@ short_description: Street image segmentation with SegFormer.
 ## Live demo
 
 The public [Hugging Face Space](https://huggingface.co/spaces/Mingze/StreetSceneSegmentation)
-runs entirely in the browser interface. Students can upload an image, choose a
-built-in example, inspect the results, and download the outputs without installing
+runs entirely in the browser interface. Students can upload an image, choose the
+UBC example, inspect the results, and download the outputs without installing
 Python or providing an API key.
 
 ### Featured classroom example: UBC campus street
 
-![UBC campus street with trees, pedestrians, and a cyclist](https://raw.githubusercontent.com/LabMingzeChen/HNIVision/main/space/examples/ubc-campus-main-mall.jpeg)
+<table>
+  <tr>
+    <th>Original image</th>
+    <th>Segmentation overlay</th>
+    <th>Cityscapes color mask</th>
+  </tr>
+  <tr>
+    <td><img src="https://raw.githubusercontent.com/LabMingzeChen/HNIVision/main/space/examples/ubc-campus-main-mall.jpeg" alt="UBC campus street input"></td>
+    <td><img src="https://raw.githubusercontent.com/LabMingzeChen/HNIVision/main/space/examples/ubc-campus-segmentation-overlay.webp" alt="UBC campus segmentation overlay"></td>
+    <td><img src="https://raw.githubusercontent.com/LabMingzeChen/HNIVision/main/space/examples/ubc-campus-segmentation-mask.webp" alt="UBC campus Cityscapes color mask"></td>
+  </tr>
+</table>
 
 This scene provides a useful teaching example because it combines vegetation,
 sky, terrain, paving, buildings, poles, pedestrians, and a cyclist in one image.
 Students can compare the predicted area shares and inspect challenging boundaries
 between the tree canopy and sky, or between the paved route and grass.
 
+The result above was generated by the live Space at 2048 × 1536 px with an
+overlay opacity of 0.55. The largest predicted classes were:
+
+| Predicted class | Image share |
+|---|---:|
+| Vegetation | 42.87% |
+| Sky | 20.86% |
+| Road | 20.73% |
+| Terrain | 11.50% |
+| Building | 1.76% |
+| Sidewalk | 1.07% |
+
 ## What the project does
 
 The application uses
@@ -67,6 +90,24 @@ For each image, the app produces:
 | Raw class-ID PNG | Preserves the numeric prediction for later analysis |
 | CSV summary | Supports statistics, plotting, and classroom exercises |
 
+## Model and dataset attribution
+
+This repository provides the Gradio application and output utilities; it did not
+train the segmentation model or create its training dataset. The external
+resources used by the app are credited below.
+
+| Borrowed resource | Creator and citation | How it is used |
+|---|---|---|
+| [SegFormer architecture](https://arxiv.org/abs/2105.15203) | Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M. Alvarez, and Ping Luo (2021) | Transformer encoder and lightweight MLP decoder architecture |
+| [`nvidia/segformer-b0-finetuned-cityscapes-1024-1024`](https://huggingface.co/nvidia/segformer-b0-finetuned-cityscapes-1024-1024) | NVIDIA checkpoint; model card maintained on Hugging Face | Public pretrained weights used for inference |
+| [Cityscapes dataset](https://www.cityscapes-dataset.com/) | Marius Cordts et al. (CVPR 2016) | Fine-tuning dataset, 19 evaluation classes, and standard class palette |
+
+The MIT license in this repository covers **only this project's original code**.
+It does not relicense the model weights or Cityscapes. The Hugging Face model
+card lists the checkpoint license as `other`, and Cityscapes is subject to its
+[official terms](https://www.cityscapes-dataset.com/license/), including
+non-commercial-use and attribution conditions. Review those terms before reuse.
+
 ## The 19 Cityscapes classes
 
 | Group | Classes | Example questions |
@@ -107,11 +148,10 @@ to limit memory use while preserving aspect ratio.
 - **Model literacy:** inspect boundary errors and discuss why predictions differ from human interpretation.
 - **Change over time:** process repeat photographs of a location and compare their CSV summaries.
 
-Five sample street scenes, led by the UBC campus example above, are loaded from the
+The UBC campus image above is the only built-in example and appears directly
+beneath the upload panel in the live app. It is hosted in the
 [HNIVision example collection](https://github.com/LabMingzeChen/HNIVision/tree/main/space/examples)
-and appear directly beneath the upload panel in the live app. Hosting samples
-through GitHub Raw follows the same lightweight pattern as HNI-VLM and avoids
-duplicating binary assets in the Space repository.
+and loaded through GitHub Raw, following the same lightweight pattern as HNI-VLM.
 
 ## Run locally
 
@@ -185,15 +225,28 @@ StreetSceneSegmentation/
 - Model-license and dataset terms remain governed by the linked model card and
   [Cityscapes](https://www.cityscapes-dataset.com/) terms.
 
-## References
+## Cite the model and dataset
+
+If you use this application in research, cite both the SegFormer architecture
+and the Cityscapes dataset:
 
-- Enze Xie et al., [“SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers”](https://arxiv.org/abs/2105.15203), NeurIPS 2021.
-- Marius Cordts et al., [“The Cityscapes Dataset for Semantic Urban Scene Understanding”](https://arxiv.org/abs/1604.01685), CVPR 2016.
-- NVIDIA, [SegFormer-B0 Cityscapes model card](https://huggingface.co/nvidia/segformer-b0-finetuned-cityscapes-1024-1024).
+```bibtex
+@article{xie2021segformer,
+  title   = {SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers},
+  author  = {Xie, Enze and Wang, Wenhai and Yu, Zhiding and Anandkumar, Anima and Alvarez, Jose M. and Luo, Ping},
+  journal = {Advances in Neural Information Processing Systems},
+  year    = {2021}
+}
 
-## Citation
+@inproceedings{Cordts2016Cityscapes,
+  title     = {The Cityscapes Dataset for Semantic Urban Scene Understanding},
+  author    = {Cordts, Marius and Omran, Mohamed and Ramos, Sebastian and Rehfeld, Timo and Enzweiler, Markus and Benenson, Rodrigo and Franke, Uwe and Roth, Stefan and Schiele, Bernt},
+  booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
+  year      = {2016}
+}
+```
 
-If this software supports your teaching or research, cite the repository:
+You may additionally cite this software:
 
 ```bibtex
 @software{chen2026streetscenesegmentation,

app.py

@@ -41,16 +41,7 @@ SAMPLE_ROOT = (
     "https://raw.githubusercontent.com/"
     "LabMingzeChen/HNIVision/main/space/examples"
 )
-SAMPLE_IMAGES = [
-    f"{SAMPLE_ROOT}/{filename}"
-    for filename in (
-        "ubc-campus-main-mall.jpeg",
-        "140546.jpg",
-        "145541.jpg",
-        "148831.jpg",
-        "160710.jpg",
-    )
-]
+SAMPLE_IMAGES = [f"{SAMPLE_ROOT}/ubc-campus-main-mall.jpeg"]
 
 
 @lru_cache(maxsize=1)
@@ -185,8 +176,8 @@ with gr.Blocks(title="Street Scene Segmentation", theme=gr.themes.Soft(), css=CS
                 gr.Examples(
                     examples=SAMPLE_IMAGES,
                     inputs=image_input,
-                    label="Or try a sample street scene",
-                    examples_per_page=4,
+                    label="Try the UBC campus street example",
+                    examples_per_page=1,
                 )
             with gr.Accordion("Display settings", open=False):
                 opacity_input = gr.Slider(