Plotting

Utilities for plotting various things

# hide
from nbdev.showdoc import *

Plotting object detection datasets

COCO

---

source

plot_coco_instance

def plot_coco_instance(
    coco_data:dict, image_id:int, image_dir:Path, ax:Axes, plot_bbox:bool=True, plot_segmentation:bool=True,
    show_labels:bool=True, show_title:bool=True
)->Axes:

Call self as a function.

import json

with open('example_data/tiles/coco_norm.json') as f:
    coco_data = json.load(f)

fig, ax = plt.subplots(1,1)
plot_coco_instance(coco_data, 1, Path('example_data/tiles/images/'), ax, plot_bbox=True)
plt.show()

Also works with oriented bounding boxes

with open('example_data/tiles/coco_rot.json') as f:
    coco_rot_data = json.load(f)
fig, ax = plt.subplots(1,1)
plot_coco_instance(coco_rot_data, 3, Path('example_data/tiles/images/'), ax, plot_bbox=True)
plt.show()

YOLO

---

source

plot_yolo_instance

def plot_yolo_instance(
    annotation_fname:Path, image_fname:Path, ax:Axes, ann_type:str='box', show_title:bool=True, classes:list=None
)->Axes:

Plot yolo format instance to ax and return it. If classes are provided show them in legend

fig, ax = plt.subplots(1,1)
classes = ['standing', 'fallen']
plot_yolo_instance(Path('example_data/tiles/labels/R0C3.txt'), Path('example_data/tiles/images/R0C3.tif'), ax, ann_type='polygon',
                   classes=classes)