openclassroom-ai/oc-p09-street-vision-poc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

docs: README bilingue FR/EN (racine + app/)

Browse source 
Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
Grégoire Mureau 2026-09-10 01:50:07 +02:00
parent 599aa3a67d
commit 18c6a1c949
4 changed files with 245 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
README.MD → README.fr.md
View file

@ -1,5 +1,7 @@
# Street Vision PoC - Segmentation Sémantique Urbaine # Street Vision PoC - Segmentation Sémantique Urbaine
📘 Ce projet est également disponible en [anglais 🇬🇧](./README.md)
**Source de vérité** : [Forgejo](https://git.gregoiremureau.com/openclassroom-ai/oc-p09-street-vision-poc) (auto-hébergé). **Source de vérité** : [Forgejo](https://git.gregoiremureau.com/openclassroom-ai/oc-p09-street-vision-poc) (auto-hébergé).
```bash ```bash

131
README.md Normal file
View file

@ -0,0 +1,131 @@
# Street Vision PoC - Urban Semantic Segmentation
📘 This project is also available in [French 🇫🇷](./README.fr.md)
**Source of truth** : [Forgejo](https://git.gregoiremureau.com/openclassroom-ai/oc-p09-street-vision-poc) (self-hosted).
```bash
git clone https://git.gregoiremureau.com/openclassroom-ai/oc-p09-street-vision-poc.git
git clone https://git.gregoiremureau.com/openclassroom-ai/oc-p09-street-vision-poc.git
```
Prod: https://api.gregoiremureau.com/ml/street-vision-poc/
Dataset: `./fetch-data.sh` (Cityscapes **not** in git).
## 🎯 Project Description
**Street Vision PoC** is a proof of concept demonstrating the improvement of urban semantic segmentation performance by comparing traditional CNN architectures (FPN) with modern Transformer approaches (SegFormer).
This project empirically evaluates the potential of **Vision Transformers** for understanding complex urban scenes, with a focus on safety-critical classes (pedestrians, vehicles, signage).
### 🔬 Technical Context
- **Dataset**: Cityscapes (2,975 high-resolution pixel-by-pixel annotated images)
- **Compared architectures**:
- **Traditional CNN**: FPN + EfficientNet-B0/ResNet34
- **Vision Transformer**: SegFormer-B0/B1
- **Evaluation resolutions**: 512×512 and 768×768 pixels
- **Main metric**: Mean Intersection over Union (mIoU)
## 📊 Main Results
| **Model** | **mIoU** | **Parameters** | **Size** | **Performance** |
|------------|----------|----------------|------------|-----------------|
| **SegFormer-B1** | **79.98%** | 13.68M | 52MB | ⭐ **Best** |
| **FPN+ResNet34** | 79.22% | 23.16M | 88MB | 🔄 Strong baseline |
| **FPN+EfficientNet-B0** | 78.67% | 5.76M | 22MB | 💡 Compact |
| **SegFormer-B0** | 77.91% | 3.72M | 14MB | ⚡ Ultra-lightweight |
**Comparative analysis**: SegFormer-B1 and FPN+ResNet34 achieve similar performance (~79-80% mIoU), but SegFormer-B1 has the advantage of being more compact and likely faster on GPU. The SegFormer-B2 to B5 versions would promise significant gains, but my hardware constraints did not allow their evaluation.
### 🎯 Significant Gains by Critical Classes
- **Human**: +1.78% IoU (safety-critical)
- **Vehicle**: +1.05% IoU (mobile obstacles)
- **Object**: +0.56% IoU (urban signage)
## 🚀 Quick Deployment
### Docker Hub
The application is containerized and available on Docker Hub:
```bash
docker pull grgmdmn/street-vision-poc_api
# live prod:
# https://api.gregoiremureau.com/ml/street-vision-poc/
```
### Configuration with Docker Compose
1. Copy the example configuration file:
```bash
cp docker-compose.yml.example docker-compose.yml
```
2. Configure the environment variables in `docker-compose.yml`.
3. Launch the application:
```bash
docker-compose up -d
```
4. Access the interface: `http://localhost:$(port)`
## 🎮 Interface Features
### Interactive Dashboard
- **Image upload**: Drag-and-drop of urban images
- **Dual inference**: Simultaneous comparison of FPN vs SegFormer
- **Visualization**: Segmentation masks colored by class
- **Real-time metrics**: Inference time and confidence scores
- **Export**: Download of segmentation results
- **WCAG accessibility**: Interface compliant with accessibility standards
### Exploratory Data Analysis
- **Cityscapes statistics**: Class distribution, native resolutions
- **Interactive charts**: Pixel breakdown by meta-class
- **Annotated examples**: Visualization of dataset samples
## 📈 Evaluation Methodology
### Comparative Approach
1. **CNN baseline**: FPN+EfficientNet/ResNet retraining at higher resolutions
2. **New approach**: SegFormer evaluation with global attention
3. **Rigorous protocol**: Cross-validation, standardized metrics
4. **Per-class analysis**: Detailed performance across 8 urban meta-classes
### Identified Limitations
- **Fixed resolution**: Evaluation without sliding window (vs. original protocol)
- **Limited resources**: Tests only on SegFormer-B0/B1
- **Controlled conditions**: Cityscapes dataset only (favorable weather)
## 🔬 Technical Insights
### SegFormer Advantages
- **Global attention**: Every pixel "sees" the entire image simultaneously
- **Lighter architecture**: Compact model with fewer parameters
- **Contextual robustness**: Better handling of occlusions and complex objects
- **Improvement potential**: B2-B5 versions would promise significant gains
### Identified Trade-offs
- **Classic CPU speed**: ~2x slower than FPN on high-performance processors (Xeon, 9600X)
- **Energy-efficient CPU speed**: 3x faster on energy-efficient processors (N100)
- **GPU optimization**: Transformer architecture favored by hardware acceleration
## 📚 References
- **Xie, E. et al.** (2021). *SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers*. NeurIPS 2021.
- **Cordts, M. et al.** (2016). *The Cityscapes Dataset for Semantic Urban Scene Understanding*. CVPR 2016.
- **Segmentation Models PyTorch** - [Implementation framework used](https://github.com/qubvel/segmentation_models.pytorch)
---
**Note**: This proof of concept demonstrates the potential of Transformer architectures for urban segmentation, while highlighting the importance of computational resources to fully leverage their capabilities.

2
app/README.MD → app/README.fr.md
View file

@ -1,6 +1,8 @@
# Dashboard de Comparaison de Modèles # Dashboard de Comparaison de Modèles
📘 Cette documentation est aussi disponible en [anglais 🇬🇧](./README.md)
## Conformité aux Exigences ## Conformité aux Exigences
### Analyse Exploratoire des Données Non Structurées (Images) ### Analyse Exploratoire des Données Non Structurées (Images)

110
app/README.md Normal file
View file

@ -0,0 +1,110 @@
# Model Comparison Dashboard
📘 This documentation is also available in [French 🇫🇷](./README.fr.md)
## Compliance with Requirements
### Exploratory Analysis of Unstructured Data (Images)
**Display of image examples by category:**
- Visualization of Cityscapes dataset images organized by 8 segmentation meta-classes
- Interactive selection of training and test images with preview
- Category navigation interface with user-friendly display names
**Data counting and distribution:**
- Pie charts detailing the distribution of the 25,000 Cityscapes images
- Visualization of project usage: 80% train / 20% validation / 100% test
- Class distribution chart for the training dataset
- Summary tables with detailed statistics per split
**Image transformations:**
- Interactive demonstration of preprocessing: resizing and normalization
- Visualization of random data augmentations (rotation, flip, brightness, etc.)
- Before/after comparison of transformations with visual alignment
- Complete documentation of the 6 possible augmentation types
### Input Data Selection
**Multiple selection interface:**
- Dropdown list for Cityscapes test images with ground truth information
- Upload of personal files (JPEG, PNG) with format validation
- Preview of selected images before processing
- Error handling and real-time user feedback
### Prediction Results
**Full model comparison:**
- Side-by-side display of predictions from the two best models
- Visualizations with original image, predicted masks, and ground truth
- Detection and highlighting of segmentation errors
- Detailed per-class statistics with percentages and pixel counts
**Performance metrics:**
- Comparative inference time with speed analysis
- Performance scores (mIoU, Accuracy) per model
- Per-class IoU for detailed performance analysis
- Interactive comparative result charts
### Cloud Deployment
**Deployed architecture:**
- FastAPI application for the model comparison API
- Streamlit interface for the user dashboard
- Deployment on cloud infrastructure with high availability
- Automatic resource management and scaling
### WCAG Accessibility
**Compliance with essential criteria:**
#### Color Contrast (WCAG AA)
- Color palette with contrast ratios ≥ 4.1:1 for graphical elements
- Contrast ratios ≥ 4.5:1 for text on background
- 8 main colors validated: blue (#0066CC), red (#CC0000), green (#006600), etc.
- Replacement of gradients with solid, high-contrast colors
#### Text Alternatives
- Detailed alternative descriptions for all automatically generated charts
- Explanatory text for each visualization (pie charts, bar charts)
- Appropriate ARIA labels for interactive elements
- Complete textual documentation of image transformations
#### Non-Color-Only Indication
- Patterns and hatching complementary to colors (optional via checkbox)
- 6 different patterns: diagonal, vertical, horizontal, dots, cross
- Differentiation by shape and texture in addition to colors
- Reinforced borders and outlines for visual distinction
#### Keyboard Navigation
- Visible CSS focus with 2px blue outlines (#0066CC)
- Full support for tab navigation
- Clear visual indicators for focused elements
- Compliance with logical navigation order
#### Accessible Data Tables
- Structured tables complementary to all charts
- Appropriate headers with scope and descriptions
- Exportable data readable by screen readers
- Responsive configuration for all screens
**Built-in assistive features:**
- "Enable patterns (visual accessibility)" checkbox for customization
- Automatic chart descriptions with complete statistics
- Interface adaptable to the needs of users with disabilities
- Complete documentation of accessibility features via the `/accessibility-info` endpoint
## Technical Architecture
**Technology stack:**
- Backend: FastAPI + PyTorch for model inference
- Frontend: Streamlit with accessible components
- Visualizations: Matplotlib with WCAG palettes
- REST API documented with Pydantic validation
- Optimized state management for model comparisons
**Performance and reliability:**
- Optimized inference times (< 1 second per image)
- Efficient memory management with automatic figure closing
- File format and size validation
- Automatic cleanup of temporary files