Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Alternate approach to cropping and resizing an image before inference #23

Merged
merged 2 commits into from
May 16, 2024
Merged

Alternate approach to cropping and resizing an image before inference #23

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5d5a3ce
alternate approach to cropping and resizing an image before inference
pleary May 13, 2024
1e07500
fix variable names, use native min method
pleary May 16, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 25 additions & 10 deletions lib/inat_inferrer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -556,17 +556,32 @@ def limit_leaf_scores_that_include_humans(self, leaf_scores):

@staticmethod
def prepare_image_for_inference(file_path):
mime_type = magic.from_file(file_path, mime=True)
# attempt to convert non jpegs
if mime_type != "image/jpeg":
im = Image.open(file_path)
image = im.convert("RGB")
else:
image = tf.io.read_file(file_path)
image = tf.image.decode_jpeg(image, channels=3)
image = Image.open(file_path)
if image.mode != "RGB":
image = image.convert("RGB")
image = tf.image.convert_image_dtype(image, tf.float32)
image = tf.image.resize_with_crop_or_pad(
image, 299, 299

eventual_size = 299
central_crop_factor = 0.875
resize_min_dimension = eventual_size / central_crop_factor

height, width = image.shape[0], image.shape[1]
resize_ratio = min(height, width) / resize_min_dimension
new_height = math.ceil(height / resize_ratio)
new_width = math.ceil(width / resize_ratio)
# resize the image so we can take a central crop without needing to resample again
image = tf.image.resize(
image,
[new_height, new_width],
method=tf.image.ResizeMethod.AREA,
preserve_aspect_ratio=True
)
# determine the upper-left corner that needs to be used to grab the square crop
offset_height = math.floor((new_height - eventual_size) / 2)
offset_width = math.floor((new_width - eventual_size) / 2)
# take a square crop out of the resized image
image = tf.image.crop_to_bounding_box(
image, offset_height, offset_width, eventual_size, eventual_size
)
return tf.expand_dims(image, 0)

Expand Down
Loading