minor changes to readme

README.md

@@ -61,7 +61,8 @@ Final quantitative comparison is reported on the locked Stanford validation spli
 | robust_hard      |   0.6839 |   0.7778 |    0.6900 |   0.8355 |    0.8600 |    0.9055 |
 | robust_occlusion |   0.6317 |   0.7317 |    0.6366 |   0.8048 |    0.8060 |    0.8600 |
 
-The augmentation-v2 EfficientNet-B3 model improved both clean validation accuracy and robustness compared with the earlier EfficientNet-B3 candidate.
+Compared with the earlier EfficientNet-B3 candidate (no augmentation v2), this model improves clean fine accuracy by +1.6 pts (0.770 → 0.786) and 
+robustness substantially — robust_hard +14 pts (0.543 → 0.684) and robust_occlusion +13 pts (0.502 → 0.632). Full comparison: the GitHub experiment report.
 
 ## Robustness evaluation
 
@@ -83,7 +84,7 @@ The main reason is that Stanford Cars and CompCars have a significant domain and
 - Stanford Cars is a clean fine-grained benchmark with 196 make/model/year classes.
 - CompCars contains different image domains and different make/model taxonomies.
 - Exact Stanford Cars ↔ CompCars make/model/year overlap was too small and biased for safe blind merging.
-- Make-level external validation on CompCars showed a strong cross-domain performance drop.
+- Make-level external validation on CompCars dropped to ~28% (vs ~0.87 make accuracy in-domain), confirming a large cross-domain shift.
 
 This confirmed that CompCars integration is a domain adaptation problem, not a simple data-merge task.
 
@@ -172,6 +173,9 @@ Appropriate uses:
 - prototype vehicle inspection workflow
 - academic/academy project demonstration
 
+Data note: weights are trained on the Stanford Cars dataset (research/educational use); the MIT license covers the project code. 
+Use of the weights should respect the Stanford Cars dataset terms.
+
 ## Limitations
 
 - The final model is trained on Stanford Cars, not on real drone/robot production footage.