Update README.md

README.md

@@ -137,7 +137,7 @@ The following table shows the original and perturbed images generated by the abo
 | Confidence: 0.44                                                     | Probability: 0.25                               |
 | ![Image 1](sample_images/imagenet/hare_ILSVRC2012_val_00004064.JPEG) | ![Image 2](output_images/hare_to_goldfish.JPEG) |
 
-As can be seen, the perturbed image is misclassified by the model as the target category `goldfish` with a confidence of `0.9999`. The perturbed image looks almost identical to the original image, demonstrating the effectiveness of the adversarial attack and the limitations of the model.
+As can be seen, the perturbed image is misclassified by the model as the target category `goldfish` with a confidence of `0.25`. The perturbed image looks almost identical to the original image, demonstrating the effectiveness of the adversarial attack and the limitations of the model.
 
 NOTE: The perturbed image is a cropped version of the original image due to the preprocessing steps needed to pass the image throught the model. The perturbed image could be upscaled using information from the original image. This is left as future work. 
 ## Testing