New feature: drop images that are too narrow or too tall

[garyfeng]

As the user I want to add an option to limit the download to only images that are neither too wide nor too tall. The option --aspect_ratio_threshold takes a float number between 0-1 as value. For example, --aspect_ratio_threshold 0.5 means that we only keep images where min(h,w)/max(h,w)>=0.5, where h,w and the height and width of the image. In this case, a 240x500 or 800x300 image would be too wide/tall to keep, but a 250x500 image will be kept.

• The offending images will be erased from the output folder, but the log entry will remain if --extract_metadata is used.
• If --coco_metadata is used, the offending image will not be shown in the coco metadata output.

Rational -- a downstream step will generate OOM error when the image is too narrow or too tall. The threshold happens to be 0.5. Rather than having an extra step to filter out these images, we can add this as a download requirement.

[garyfeng]

This does it

googleimagesdownload -k "swimming free style" -oc "male swimmer" -l 50 --extract_metadata --coco_metadata --type photo --size ">640*480" --format jpg --aspect_ratio_threshold 0.5 --output_directory "E:\swimpose" --image_directory "square"

and the output will contain

Deleting Image, aspect ratio 0.56 ====> 20.n2kia8vhwhbeiuyyzb0u.jpg
Deleting Image, aspect ratio 0.56 ====> 22.m1m2018c016.jpg