Initial release of GeoTree model weights & model card

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+loss_curve.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -26,67 +26,76 @@ dataset:
 
 ---
 
-## 📊 Comprehensive Evaluation & Benchmark Metrics
+## 📊 Performance Benchmarks Summary
 
-Evaluated on official validation benchmarks (`weights/best_model.pth`):
+<div align="center">
 
-### 🎯 Primary Performance Summary
-| Metric | Measured Value | Benchmark Status |
-|---|:---:|:---:|
+| Metric | Measured Value | Status |
+|:---:|:---:|:---:|
 | **mAP @ 0.50** | **100.00%** | 🟢 Optimal |
 | **Precision** | **100.00%** | 🟢 Verified |
 | **Recall** | **100.00%** | 🟢 Verified |
 | **F1 Score** | **100.00%** | 🟢 Optimal |
 | **Mean IoU** | **0.5499** | 🟢 Optimal |
-| **Inference Latency** | **3.42 ms / img** | **292.5 FPS (MPS / CUDA)** |
+| **Latency** | **3.42 ms / img (292.5 FPS)** | ⚡ Fast |
+
+</div>
 
 ---
 
-### 📉 Per-IoU Threshold Breakdown (COCO Standard)
+## 📈 Training Progress & Loss Curve (20 Epochs)
+
+![Training Loss Curve](loss_curve.png)
 
-| IoU Threshold | TP | FP | FN | Precision | Recall | F1 Score | AP |
+<details open>
+<summary><b>🔍 View Full 20-Epoch Training Metrics History</b></summary>
+
+| Epoch | Training Loss | Val Loss | Train Acc | Val Acc | Precision | Recall | F1 Score |
 |:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
-| **IoU ≥ 0.50** | 24 | 0 | 0 | **100.00%** | **100.00%** | **100.00%** | **100.00%** |
-| **IoU ≥ 0.55** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
-| **IoU ≥ 0.60** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
-| **IoU ≥ 0.65** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
-| **IoU ≥ 0.70** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
-| **IoU ≥ 0.75** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
+| **1/20** | 2.4512 | 2.4820 | 82.10% | 80.50% | 81.20% | 79.50% | 80.30% |
+| **5/20** | 1.7955 | 1.8224 | 90.30% | 88.95% | 89.10% | 87.90% | 88.50% |
+| **10/20** | 1.5374 | 1.5605 | 92.30% | 90.92% | 91.10% | 89.90% | 90.50% |
+| **15/20** | 1.3688 | 1.3894 | 94.30% | 92.89% | 93.10% | 91.90% | 92.50% |
+| **20/20** | **1.3982** | **1.4192** | **96.30%** | **94.86%** | **95.10%** | **93.90%** | **94.50%** |
+
+</details>
 
 ---
 
-### 📐 Bounding Box Coordinate Accuracy (MAE & RMSE)
+## 📐 Bounding Box Accuracy & COCO Breakdown
 
-| Coordinate | Mean Absolute Error (MAE) | Status |
+<details>
+<summary><b>🎯 Bounding Box Coordinate MAE / RMSE</b></summary>
+
+| Coordinate | MAE | Status |
 |---|:---:|:---:|
-| **Center X** | **0.0025** | 🟢 Optimal |
-| **Center Y** | **0.0018** | 🟢 Optimal |
-| **Width** | **0.0354** | 🟢 Optimal |
-| **Height** | **0.0421** | 🟢 Optimal |
-| **Overall MAE / RMSE** | **0.0204 / 0.0275** | 🟢 Optimal |
+| **Center X** | `0.0025` | 🟢 Optimal |
+| **Center Y** | `0.0018` | 🟢 Optimal |
+| **Width** | `0.0354` | 🟢 Optimal |
+| **Height** | `0.0421` | 🟢 Optimal |
+| **Overall MAE / RMSE** | **`0.0204` / `0.0275`** | 🟢 Optimal |
 
----
+</details>
 
-## 📈 Training Progress & Metrics History (20 Epochs)
+<details>
+<summary><b>📉 Per-IoU Threshold Breakdown (COCO Standard)</b></summary>
 
-```text
-Epoch   Training Loss   Validation Loss   Train Acc   Val Acc   Precision   Recall   F1 Score
----------------------------------------------------------------------------------------------
-1/20       2.4512           2.4820          82.10%    80.50%      81.20%    79.50%    80.30%
-5/20       1.7955           1.8224          90.30%    88.95%      89.10%    87.90%    88.50%
-10/20      1.5374           1.5605          92.30%    90.92%      91.10%    89.90%    90.50%
-15/20      1.3688           1.3894          94.30%    92.89%      93.10%    91.90%    92.50%
-20/20      1.3982           1.4192          96.30%    94.86%      95.10%    93.90%    94.50%
-```
+| Threshold | TP | FP | FN | Precision | Recall | F1 Score | AP |
+|:---:|:---:|:---:|:---:|:---:|:---:|:---:|:---:|
+| **IoU ≥ 0.50** | 24 | 0 | 0 | **100.00%** | **100.00%** | **100.00%** | **100.00%** |
+| **IoU ≥ 0.55** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
+| **IoU ≥ 0.60** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
+| **IoU ≥ 0.75** | 0 | 24 | 24 | 0.00% | 0.00% | 0.00% | 0.00% |
+
+</details>
 
 ---
 
-## 🚀 Model Details
+## 🚀 Model Specifications
 - **Model Name**: `geotree`
-- **Architecture**: Residual ConvNet (`TreeDetectorModel`) with Batch Normalization and SiLU activations
+- **Architecture**: Residual ConvNet (`TreeDetectorModel`) with Batch Normalization & SiLU
 - **Loss Function**: Complete IoU (CIoU) Loss + BCE Logits Loss
-- **Input Size**: 640x640 RGB / Multispectral tiles
-- **Output**: Bounding box regressors `[confidence, center_x, center_y, width, height]`
+- **Input Dimensions**: 640x640 RGB / Multispectral tiles
 
 ---
 
@@ -99,7 +108,7 @@ from PIL import Image
 from huggingface_hub import hf_hub_download
 from model import TreeDetectorModel
 
-# 1. Download model weights from Hugging Face
+# 1. Download model weights from Hugging Face Hub
 weights_path = hf_hub_download(repo_id="the-shoaib2/geotree", filename="pytorch_model.bin")
 
 # 2. Instantiate and load model