Update README.md

README.md

@@ -8,45 +8,42 @@ pinned: false
 ---
 
 
-# Random Baseline Model for Climate Disinformation Classification
+# Object Detection Model for Smoke detection
 
 ## Model Description
 
-This is a random baseline model for the Frugal AI Challenge 2024, specifically for the text classification task of identifying climate disinformation. The model serves as a performance floor, randomly assigning labels to text inputs without any learning.
+This is a YOLO11s model developed for the Frugal AI Challenge 2025, specifically for the object detection task of identifying smoke.
+Training was conducted over 100 epochs using custom hyperparameters and augmented data.
+We employed a post-training optimization phase that included an optimized engine (TensorRT), pruning, and quantization techniques to compress the model and reduce its memory footprint.
+The objective was to achieve the best possible accuracy while minimizing energy consumption.
 
 ### Intended Use
 
-- **Primary intended uses**: Baseline comparison for climate disinformation classification models
-- **Primary intended users**: Researchers and developers participating in the Frugal AI Challenge
-- **Out-of-scope use cases**: Not intended for production use or real-world classification tasks
+- **Primary intended uses**: Detect smoke based on images from watchtowers cameras.
+- **Primary intended users**: Firefighters to tackle early wildfires developments
+- **Out-of-scope use cases**: Not intended for production use or real-world object detection tasks
 
 ## Training Data
+The model uses the Pyro-SDIS is a dataset designed for wildfire smoke detection using AI models. 
+It is developed in collaboration with the Fire and Rescue Services (SDIS) in France and the dedicated volunteers of the Pyronear association.
+- Size : Training set : 29537 - Validation set : 4099
+- 1 category of objects to detect : Smoke
 
-The model uses the QuotaClimat/frugalaichallenge-text-train dataset:
-- Size: ~6000 examples
-- Split: 80% train, 20% test
-- 8 categories of climate disinformation claims
-
-### Labels
-0. No relevant claim detected
-1. Global warming is not happening
-2. Not caused by humans
-3. Not bad or beneficial
-4. Solutions harmful/unnecessary
-5. Science is unreliable
-6. Proponents are biased
-7. Fossil fuels are needed
+### Objects
+0. Smoke
 
 ## Performance
 
-### Metrics
-- **Accuracy**: ~12.5% (random chance with 8 classes)
-- **Environmental Impact**:
-  - Emissions tracked in gCO2eq
-  - Energy consumption tracked in Wh
+### Metrics on Validation Dataset
+- **Accuracy**: ~80%% 
+- **Inference Environmental Impact**:
+  - for 4099 images on a Tesla T4 GPU 
+  - 1.3 gCO2 // Emissions tracked in gCO2eq
+  - 3.5 Wh // Energy consumption tracked in Wh
 
 ### Model Architecture
-The model implements a random choice between the 8 possible labels, serving as the simplest possible baseline.
+The model implements a YOLO11s with augmented data using image processing techniques such as zooming, adding noise and contrast adjustment. 
+The model was then optimized for high performance inference using TensorRT, pruning and quantization technics to compress and limit model memory footprint.
 
 ## Environmental Impact
 
@@ -57,15 +54,10 @@ Environmental impact is tracked using CodeCarbon, measuring:
 This tracking helps establish a baseline for the environmental impact of model deployment and inference.
 
 ## Limitations
-- Makes completely random predictions
-- No learning or pattern recognition
-- No consideration of input text
-- Serves only as a baseline reference
-- Not suitable for any real-world applications
+- Not suitable for industrialization or real-world applications
 
 ## Ethical Considerations
 
-- Dataset contains sensitive topics related to climate disinformation
-- Model makes random predictions and should not be used for actual classification
+- Model makes predictions and should be used with caution, can be false predictions.
 - Environmental impact is tracked to promote awareness of AI's carbon footprint
-```
+```