!mkdir _temp
!curl http://farm1.static.flickr.com/220/512450093_7717fb8ce8.jpg > _temp/cat.jpg
!echo `pwd`/_temp/cat.jpg > _temp/cat.txt
!../python/detect.py --crop_mode=selective_search --pretrained_model=imagenet/caffe_reference_imagenet_model --model_def=imagenet/imagenet_deploy.prototxt _temp/cat.txt _temp/cat.h5

import pandas as pd

df = pd.read_hdf('_temp/cat.h5', 'df')
print(df.shape)
print(df.iloc[0])

with open('../data/ilsvrc12/synset_words.txt') as f:
    labels_df = pd.DataFrame([
        {
            'synset_id': l.strip().split(' ')[0],
            'name': ' '.join(l.strip().split(' ')[1:]).split(',')[0]
        }
        for l in f.readlines()
    ])
labels_df.sort('synset_id')
predictions_df = pd.DataFrame(np.vstack(df.prediction.values), columns=labels_df['name'])
print(predictions_df.iloc[0])

gray()
matshow(predictions_df.values)
xlabel('Classes')
ylabel('Windows')

max_s = predictions_df.max(0)
max_s.sort(ascending=False)
print(max_s[:10])

# Find, print, and display max detection.
window_order = pd.Series(predictions_df.values.max(1)).order(ascending=False)

i = window_order.index[3]
j = window_order.index[13]

# Show top predictions for top detection.
f = pd.Series(df['prediction'].iloc[i], index=labels_df['name'])
print('Top detection:')
print(f.order(ascending=False)[:5])
print('')

# Show top predictions for 10th top detection.
f = pd.Series(df['prediction'].iloc[j], index=labels_df['name'])
print('10th detection:')
print(f.order(ascending=False)[:5])

# Show top detection in red, 10th top detection in blue.
im = imread('_temp/cat.jpg')
imshow(im)
currentAxis = plt.gca()

det = df.iloc[i]
coords = (det['xmin'], det['ymin']), det['xmax'] - det['xmin'], det['ymax'] - det['ymin']
currentAxis.add_patch(Rectangle(*coords, fill=False, edgecolor='r', linewidth=5))

det = df.iloc[j]
coords = (det['xmin'], det['ymin']), det['xmax'] - det['xmin'], det['ymax'] - det['ymin']
currentAxis.add_patch(Rectangle(*coords, fill=False, edgecolor='b', linewidth=5))

def nms_detections(dets, overlap=0.5):
    """
    Non-maximum suppression: Greedily select high-scoring detections and
    skip detections that are significantly covered by a previously
    selected detection.

    This version is translated from Matlab code by Tomasz Malisiewicz,
    who sped up Pedro Felzenszwalb's code.

    Parameters
    ----------
    dets: ndarray
        each row is ['xmin', 'ymin', 'xmax', 'ymax', 'score']
    overlap: float
        minimum overlap ratio (0.5 default)

    Output
    ------
    dets: ndarray
        remaining after suppression.
    """
    if np.shape(dets)[0] < 1:
        return dets

    x1 = dets[:, 0]
    y1 = dets[:, 1]
    x2 = dets[:, 2]
    y2 = dets[:, 3]

    w = x2 - x1
    h = y2 - y1
    area = w * h

    s = dets[:, 4]
    ind = np.argsort(s)

    pick = []
    counter = 0
    while len(ind) > 0:
        last = len(ind) - 1
        i = ind[last]
        pick.append(i)
        counter += 1

        xx1 = np.maximum(x1[i], x1[ind[:last]])
        yy1 = np.maximum(y1[i], y1[ind[:last]])
        xx2 = np.minimum(x2[i], x2[ind[:last]])
        yy2 = np.minimum(y2[i], y2[ind[:last]])

        w = np.maximum(0., xx2 - xx1 + 1)
        h = np.maximum(0., yy2 - yy1 + 1)

        o = w * h / area[ind[:last]]

        to_delete = np.concatenate(
            (np.nonzero(o > overlap)[0], np.array([last])))
        ind = np.delete(ind, to_delete)

    return dets[pick, :]

scores = predictions_df['tiger cat']
windows = df[['xmin', 'ymin', 'xmax', 'ymax']].values
dets = np.hstack((windows, scores[:, np.newaxis]))
nms_dets = nms_detections(dets)

imshow(im)
currentAxis = plt.gca()
colors = ['r', 'b', 'y']
for c, det in zip(colors, nms_dets[:3]):
    currentAxis.add_patch(
        Rectangle((det[0], det[1]), det[2], det[3],
        fill=False, edgecolor=c, linewidth=5)
    )

import shutil
shutil.rmtree('_temp')