| import os |
| import cv2 |
| import numpy as np |
| from editboard.test_optflow import compute_optical_flow, apply_optical_flow |
| from editboard.utils import load_json |
| from tqdm import tqdm |
|
|
| def get_optical_flow_list(video_path): |
| flow_list = [] |
| frames = os.listdir(video_path) |
| frames = [img for img in frames if (img.endswith('.png') or img.endswith('.jpg') or img.endswith('.jpeg'))] |
| frames.sort() |
| for i in range(0,len(frames)-1): |
| img1 = cv2.imread(os.path.join(video_path, frames[i])) |
| img2 = cv2.imread(os.path.join(video_path, frames[i+1])) |
| flow = compute_optical_flow(img1,img2) |
| flow_list.append(flow) |
| return flow_list |
|
|
| def get_warped_result_list(video_path, flow_list): |
| warp_list = [] |
| frames = os.listdir(video_path) |
| frames = [img for img in frames if (img.endswith('.png') or img.endswith('.jpg') or img.endswith('.jpeg'))] |
| frames.sort() |
| for i in range(0,len(frames)-1): |
| pp = os.path.join(video_path, frames[i]) |
| img1 = cv2.imread(pp) |
| flow = flow_list[i] |
| warped = apply_optical_flow(img1, flow) |
| warp_list.append(warped) |
| return warp_list |
|
|
| def calculate_ff_alpha(original,ori_warp,edit,edit_warp,threshold=5): |
| m,n,_ = original.shape |
| mask = np.zeros((m,n)) |
|
|
| diff = cv2.absdiff(original, ori_warp) |
| diff_gray = cv2.cvtColor(diff, cv2.COLOR_BGR2GRAY) |
|
|
| diff_edit = cv2.absdiff(edit, edit_warp) |
| |
| diff_gray_edit = np.max(diff_edit,-1) |
| for i in range(m): |
| for j in range(n): |
| if diff_gray[i][j] <= threshold: |
| mask[i][j] = 1 |
| else: |
| mask[i][j] = 0 |
|
|
| percentage_of_valid_pixel = np.sum(mask==1)/512/512 |
|
|
| a = np.sum(np.multiply(mask,diff_gray_edit)) |
| result = a/np.sum(mask==1) |
| return result, percentage_of_valid_pixel |
|
|
|
|
| def ff_alpha_for_video(original_video_path, edited_video_path, threshold = 5): |
| result = [] |
| valid_percentage = [] |
| original_frames = os.listdir(original_video_path) |
| original_frames = [img for img in original_frames if (img.endswith('.png') or img.endswith('.jpg') or img.endswith('.jpeg'))] |
| original_frames.sort() |
|
|
| edited_frames = os.listdir(edited_video_path) |
| edited_frames = [img for img in edited_frames if (img.endswith('.png') or img.endswith('.jpg') or img.endswith('.jpeg'))] |
| edited_frames.sort() |
|
|
| flow_list = get_optical_flow_list(original_video_path) |
| edit_warp_result = get_warped_result_list(edited_video_path,flow_list) |
| original_warp_result = get_warped_result_list(original_video_path,flow_list) |
|
|
| for i in range(0, len(edit_warp_result)): |
| original = cv2.imread(os.path.join(original_video_path,original_frames[i+1])) |
| ori_warp = original_warp_result[i] |
| edit = cv2.imread(os.path.join(edited_video_path,edited_frames[i+1])) |
| edit_warp = edit_warp_result[i] |
| score, valid = calculate_ff_alpha(original, ori_warp, edit, edit_warp,threshold) |
| result.append(score) |
| valid_percentage.append(valid) |
|
|
| if sum(valid_percentage)/len(valid_percentage) >= 0.70: |
| return sum(result)/len(edit_warp_result) |
| else: |
| return 0 |
|
|
| def compute_ff_alpha(json_dir, device, submodules_list): |
| metadata = load_json(json_dir) |
| result = {} |
| for i in tqdm(metadata): |
| score = ff_alpha_for_video(i["original_video_path"], i["edited_video_path"]) |
| result[i["original_video_path"] + i["edited_video_path"]] = score |
| return result |
| |
