Project 3 Predictions

Image 1:

• Actual: ~10
• Predicted: ~24

Image 2:

• Actual: 0
• Predicted: ~0

Image 3:

• Actual: ~23
• Predicted: ~14

Conclusion

As can be seen, my hypothesis that the labels were mismatched with the data was correct. There is no trend amongst the actual: predicted ratio and the actual labels seem wrong. It is clear that this heavily impacted my model; although image 2 was technically predicted correctly, it seems more of a fluke. It is hardly believable that image 1 and 3 truly have that many people.