Imagga Auto-Categorization - correct

Imagga Auto-Categorization - Modified in photoshop (blur, fragment, paint strokes, add gausian noise, texturizer) - still correct

Imagga Auto Categorization - Modified in photoshop (blur, fragment, paint strokes, add gausian noise, texturizer, selective color, hue and saturation) and then cropped - mostly incorrect.

PSENet - Text Detection - Some text correctly detected, some not.

Original - photo before Strongin house hot ones study break

MiDaSv2 - Depth estimation

Transparent green lines added in photoshop

MiDaSv2 - Depth estimation - bottles stand out less, but sticker stands out more

Solid green lines added in photoshop

MiDaSv2 - Depth estimation - bottles all but disappear, sticker still stands out but not as defined

YOLOv4 - Object classification - child classified as fire hydrant and teddy bear


Imagga Auto-Categorization - incorrect

Ximilar Dominant Colors - color extraction

Original


DenseCap - Some accurate sentences:

a green bird, the nose is black, the hand of a person

Original

DenseCap - No accurate sentences

Original

DenseCap - Some accurate, partially accurate, and inaccurate sentences:


DenseCap - Modified in photoshop - colorized and brightened - Surprisingly more accurate than the original

Dense Cap - rotated 180 degrees - Some accurate, some inaccurate. "Man is holding a cake" haha.

DenseCap - Cropping had a predictable effect on the outcome

DenseCap - "The grass is green"