avcv 0.0.1

Python package utils for computer vision

AVCV

Optimized functions for vision problems

%load_ext autoreload
%autoreload 2

from nbdev.showdoc import *

This file will become your README and also the index of your documentation.

Install

pip install avcv

How to use

plot_images[source]

plot_images(images, labels=None, cls_true=None, cls_pred=None, space=(0.3, 0.3), mxn=None, size=(5, 5), dpi=300, max_w=1500, out_file=None, cmap='binary')

Plot images

from avcv.visualize import plot_images
from glob import glob
import numpy as np
import mmcv
paths = glob('/data/synthetic/SHARE_SVA_DATASET/val/000/frames/*')
imgs = [mmcv.imread(path, channel_order='rgb') for path in np.random.choice(paths, 10)]
plot_images(imgs)

---------------------------------------------------------------------------

ModuleNotFoundError                       Traceback (most recent call last)

<ipython-input-4-c232fb9b8e39> in <module>
----> 1 from avcv.plot_images import plot_images
      2 from glob import glob
      3 import numpy as np
      4 import mmcv
      5 paths = glob('/data/synthetic/SHARE_SVA_DATASET/val/000/frames/*')


ModuleNotFoundError: No module named 'avcv.plot_images'

Multi thread

Elementwise multithreading a given function, the results are store in an array coresponding to the inputs

# example
from glob import glob
import mmcv
import numpy as np
from avcv.process import multi_thread
from tqdm import tqdm

paths = glob('/data/synthetic/SHARE_SVA_DATASET/val/000/frames/*')
def f(x):
    mmcv.imread(x, channel_order='rgb')
    return None

inputs = np.random.choice(paths, 100)
fast_imgs = multi_thread(f, inputs)

slow_imgs = [f(_) for _ in tqdm(inputs)]