lavis/common/annotator/openpose/util.py

import math
import numpy as np
import matplotlib
import cv2


def padRightDownCorner(img, stride, padValue):
    h = img.shape[0]
    w = img.shape[1]

    pad = 4 * [None]
    pad[0] = 0 # up
    pad[1] = 0 # left
    pad[2] = 0 if (h % stride == 0) else stride - (h % stride) # down
    pad[3] = 0 if (w % stride == 0) else stride - (w % stride) # right

    img_padded = img
    pad_up = np.tile(img_padded[0:1, :, :]*0 + padValue, (pad[0], 1, 1))
    img_padded = np.concatenate((pad_up, img_padded), axis=0)
    pad_left = np.tile(img_padded[:, 0:1, :]*0 + padValue, (1, pad[1], 1))
    img_padded = np.concatenate((pad_left, img_padded), axis=1)
    pad_down = np.tile(img_padded[-2:-1, :, :]*0 + padValue, (pad[2], 1, 1))
    img_padded = np.concatenate((img_padded, pad_down), axis=0)
    pad_right = np.tile(img_padded[:, -2:-1, :]*0 + padValue, (1, pad[3], 1))
    img_padded = np.concatenate((img_padded, pad_right), axis=1)

    return img_padded, pad

# transfer caffe model to pytorch which will match the layer name
def transfer(model, model_weights):
    transfered_model_weights = {}
    for weights_name in model.state_dict().keys():
        transfered_model_weights[weights_name] = model_weights['.'.join(weights_name.split('.')[1:])]
    return transfered_model_weights

# draw the body keypoint and lims
def draw_bodypose(canvas, candidate, subset):
    stickwidth = 4
    limbSeq = [[2, 3], [2, 6], [3, 4], [4, 5], [6, 7], [7, 8], [2, 9], [9, 10], \
               [10, 11], [2, 12], [12, 13], [13, 14], [2, 1], [1, 15], [15, 17], \
               [1, 16], [16, 18], [3, 17], [6, 18]]

    colors = [[255, 0, 0], [255, 85, 0], [255, 170, 0], [255, 255, 0], [170, 255, 0], [85, 255, 0], [0, 255, 0], \
              [0, 255, 85], [0, 255, 170], [0, 255, 255], [0, 170, 255], [0, 85, 255], [0, 0, 255], [85, 0, 255], \
              [170, 0, 255], [255, 0, 255], [255, 0, 170], [255, 0, 85]]
    for i in range(18):
        for n in range(len(subset)):
            index = int(subset[n][i])
            if index == -1:
                continue
            x, y = candidate[index][0:2]
            cv2.circle(canvas, (int(x), int(y)), 4, colors[i], thickness=-1)
    for i in range(17):
        for n in range(len(subset)):
            index = subset[n][np.array(limbSeq[i]) - 1]
            if -1 in index:
                continue
            cur_canvas = canvas.copy()
            Y = candidate[index.astype(int), 0]
            X = candidate[index.astype(int), 1]
            mX = np.mean(X)
            mY = np.mean(Y)
            length = ((X[0] - X[1]) ** 2 + (Y[0] - Y[1]) ** 2) ** 0.5
            angle = math.degrees(math.atan2(X[0] - X[1], Y[0] - Y[1]))
            polygon = cv2.ellipse2Poly((int(mY), int(mX)), (int(length / 2), stickwidth), int(angle), 0, 360, 1)
            cv2.fillConvexPoly(cur_canvas, polygon, colors[i])
            canvas = cv2.addWeighted(canvas, 0.4, cur_canvas, 0.6, 0)
    # plt.imsave("preview.jpg", canvas[:, :, [2, 1, 0]])
    # plt.imshow(canvas[:, :, [2, 1, 0]])
    return canvas


# image drawed by opencv is not good.
def draw_handpose(canvas, all_hand_peaks, show_number=False):
    edges = [[0, 1], [1, 2], [2, 3], [3, 4], [0, 5], [5, 6], [6, 7], [7, 8], [0, 9], [9, 10], \
             [10, 11], [11, 12], [0, 13], [13, 14], [14, 15], [15, 16], [0, 17], [17, 18], [18, 19], [19, 20]]

    for peaks in all_hand_peaks:
        for ie, e in enumerate(edges):
            if np.sum(np.all(peaks[e], axis=1)==0)==0:
                x1, y1 = peaks[e[0]]
                x2, y2 = peaks[e[1]]
                cv2.line(canvas, (x1, y1), (x2, y2), matplotlib.colors.hsv_to_rgb([ie/float(len(edges)), 1.0, 1.0])*255, thickness=2)

        for i, keyponit in enumerate(peaks):
            x, y = keyponit
            cv2.circle(canvas, (x, y), 4, (0, 0, 255), thickness=-1)
            if show_number:
                cv2.putText(canvas, str(i), (x, y), cv2.FONT_HERSHEY_SIMPLEX, 0.3, (0, 0, 0), lineType=cv2.LINE_AA)
    return canvas

# detect hand according to body pose keypoints
# please refer to https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/src/openpose/hand/handDetector.cpp
def handDetect(candidate, subset, oriImg):
    # right hand: wrist 4, elbow 3, shoulder 2
    # left hand: wrist 7, elbow 6, shoulder 5
    ratioWristElbow = 0.33
    detect_result = []
    image_height, image_width = oriImg.shape[0:2]
    for person in subset.astype(int):
        # if any of three not detected
        has_left = np.sum(person[[5, 6, 7]] == -1) == 0
        has_right = np.sum(person[[2, 3, 4]] == -1) == 0
        if not (has_left or has_right):
            continue
        hands = []
        #left hand
        if has_left:
            left_shoulder_index, left_elbow_index, left_wrist_index = person[[5, 6, 7]]
            x1, y1 = candidate[left_shoulder_index][:2]
            x2, y2 = candidate[left_elbow_index][:2]
            x3, y3 = candidate[left_wrist_index][:2]
            hands.append([x1, y1, x2, y2, x3, y3, True])
        # right hand
        if has_right:
            right_shoulder_index, right_elbow_index, right_wrist_index = person[[2, 3, 4]]
            x1, y1 = candidate[right_shoulder_index][:2]
            x2, y2 = candidate[right_elbow_index][:2]
            x3, y3 = candidate[right_wrist_index][:2]
            hands.append([x1, y1, x2, y2, x3, y3, False])

        for x1, y1, x2, y2, x3, y3, is_left in hands:
            # pos_hand = pos_wrist + ratio * (pos_wrist - pos_elbox) = (1 + ratio) * pos_wrist - ratio * pos_elbox
            # handRectangle.x = posePtr[wrist*3] + ratioWristElbow * (posePtr[wrist*3] - posePtr[elbow*3]);
            # handRectangle.y = posePtr[wrist*3+1] + ratioWristElbow * (posePtr[wrist*3+1] - posePtr[elbow*3+1]);
            # const auto distanceWristElbow = getDistance(poseKeypoints, person, wrist, elbow);
            # const auto distanceElbowShoulder = getDistance(poseKeypoints, person, elbow, shoulder);
            # handRectangle.width = 1.5f * fastMax(distanceWristElbow, 0.9f * distanceElbowShoulder);
            x = x3 + ratioWristElbow * (x3 - x2)
            y = y3 + ratioWristElbow * (y3 - y2)
            distanceWristElbow = math.sqrt((x3 - x2) ** 2 + (y3 - y2) ** 2)
            distanceElbowShoulder = math.sqrt((x2 - x1) ** 2 + (y2 - y1) ** 2)
            width = 1.5 * max(distanceWristElbow, 0.9 * distanceElbowShoulder)
            # x-y refers to the center --> offset to topLeft point
            # handRectangle.x -= handRectangle.width / 2.f;
            # handRectangle.y -= handRectangle.height / 2.f;
            x -= width / 2
            y -= width / 2  # width = height
            # overflow the image
            if x < 0: x = 0
            if y < 0: y = 0
            width1 = width
            width2 = width
            if x + width > image_width: width1 = image_width - x
            if y + width > image_height: width2 = image_height - y
            width = min(width1, width2)
            # the max hand box value is 20 pixels
            if width >= 20:
                detect_result.append([int(x), int(y), int(width), is_left])

    '''
    return value: [[x, y, w, True if left hand else False]].
    width=height since the network require squared input.
    x, y is the coordinate of top left 
    '''
    return detect_result

# get max index of 2d array
def npmax(array):
    arrayindex = array.argmax(1)
    arrayvalue = array.max(1)
    i = arrayvalue.argmax()
    j = arrayindex[i]
    return i, j