# coding=utf-8 """Given the person trajectory, get the vehicle points the """ import argparse import os import operator #import cPickle as pickle import pickle import yaml import numpy as np from glob import glob from tqdm import tqdm from visualize_real_data import load_traj from combine_traj import get_world_coordinates parser = argparse.ArgumentParser() parser.add_argument("traj_path", help="path to pedestrian dataset") parser.add_argument("anno_path", help="yaml path") parser.add_argument("h_path", help="path to homography matrix") parser.add_argument("out_path") # For running parallel jobs, set --job 4 --curJob k, where k=1/2/3/4 parser.add_argument("--job", type=int, default=1, help="total job") parser.add_argument("--curJob", type=int, default=1, help="this script run job Num") actev_scene2imgsize = { "0002": (1280.0, 720.0), "0000": (1920.0, 1080.0), "0400": (1920.0, 1080.0), "0401": (1920.0, 1080.0), "0500": (1920.0, 1080.0), } scene2imgsize = actev_scene2imgsize def load_yml_file_without_meta(yml_file): """Load the ActEV YAML annotation files.""" with open(yml_file, "r") as f: data = yaml.load(f, Loader=yaml.FullLoader) # get the meta index first mi = -1 for i in range(len(data)): if "meta" not in data[i]: mi = i break assert mi >= 0 return data[mi:] def load_tracks(track_file, only=None): """Load track object type information.""" trackid2object_ = {} data = load_yml_file_without_meta(track_file) for one in data: one = one["types"] # added in v1_update # v4, changed to - { types: { id1: 0 , cset3: { Person: 1.0 } } } if "obj_type" not in one: one["obj_type"] = list(one["cset3"].keys())[0] assert len(one["cset3"].keys()) == 1 if only is not None: if one["obj_type"] != only: continue trackid2object_[int(one["id1"])] = one["obj_type"] return trackid2object_ def load_boxes(box_file_, imgsize_): """Load bounding boxes.""" boxes_ = [] data = load_yml_file_without_meta(box_file_) for one in data: one = one["geom"] # added in v1_update trackid_ = int(one["id1"]) frame_index_ = int(one["ts0"]) bbox_ = [float(a) for a in one["g0"].split()] src = one["src"] assert src == "truth", (src, one) # check box valid is_valid = valid_box(bbox_, imgsize_) if not is_valid: # modify box to be valid? bbox_ = modify_box(bbox_, imgsize_) assert valid_box(bbox_, imgsize_) # so box is [x1, y1, x2, y2] boxes_.append((trackid_, frame_index_, bbox_)) return boxes_ def get_scene(videoname): """ActEV scene extractor from videoname.""" s = videoname.split("_S_")[-1] s = s.split("_")[0] return s[:4] # actev boxes may contain some errors # won't fix x,y reversed def modify_box(bbox, imgsize): """Modify ActEV boxes.""" w, h = imgsize x1, y1, x2, y2 = bbox x_min = min(x1, x2) x_max = max(x1, x2) y_min = min(y1, y2) y_max = max(y1, y2) x_min = min(w, x_min) x_max = min(w, x_max) y_min = min(h, y_min) y_max = min(h, y_max) return [x_min, y_min, x_max, y_max] def valid_box(box, wh): """Check whether boxes are within the image bounds.""" w, h = wh a = box_area(box) if a <= 0: return False if (box[0] > w) or (box[2] > w) or (box[1] > h) or (box[3] > h): return False return True def box_area(box): """compute bbox area size in pixels.""" x1, y1, x2, y2 = box w = x2 - x1 h = y2 - y1 return float(w) * h def get_box_center(box): x1, y1, x2, y2 = box return [(x1+x2)/2.0, (y1+y2)/2.0] def save_file(data, out_path, videoname): target_file = os.path.join(out_path, "%s.txt" % videoname) with open(target_file, "w") as f: for one in data: f.writelines("%s\n" % "\t".join(["%s" % x for x in one])) if __name__ == "__main__": args = parser.parse_args() traj_files = glob(os.path.join(args.traj_path, "*.txt")) traj_files.sort() out_path_pixel = os.path.join(args.out_path, "pixel") out_path_world = os.path.join(args.out_path, "world") if not os.path.exists(out_path_pixel): os.makedirs(out_path_pixel) os.makedirs(out_path_world) # load the homography matrices h_dict = {} h_files = glob(os.path.join(args.h_path, "*.txt")) for h_file in h_files: scene = os.path.splitext(os.path.basename(h_file))[0] h_matrix = [] with open(h_file, "r") as f: for line in f: h_matrix.append(line.strip().split(",")) h_matrix = np.array(h_matrix, dtype="float") h_dict[scene] = h_matrix count = 0 for traj_file in tqdm(traj_files): count += 1 if (count % args.job) != (args.curJob - 1): continue # 1. load pedestrian dataset, so we know what frames to look at ped_trajs = load_traj(traj_file) frame_ids = np.unique(ped_trajs[:, 0]).tolist() frame_ids.sort() frame_ids = {int(frame_id):1 for frame_id in frame_ids} videoname = os.path.splitext(os.path.basename(traj_file))[0] scene = get_scene(videoname) h_matrix = h_dict[scene] imgsize = scene2imgsize[scene] box_file = os.path.join(args.anno_path, videoname + ".geom.yml") type_file = os.path.join(args.anno_path, videoname + ".types.yml") act_file = os.path.join(args.anno_path, videoname + ".activities.yml") # load each track id and its trajectories trackid2object = load_tracks(type_file, "Vehicle") # no cars in this video? # load traj boxes for the trackid boxes = load_boxes(box_file, imgsize) data_pixel = [] # frame_id, track_id, x, y data_world = [] for box in boxes: track_id, frame_index, bbox = box #if trackid2object.has_key(track_id) and frame_ids.has_key(frame_index): if track_id in trackid2object and frame_index in frame_ids: center_xy = get_box_center(bbox) world_xy = get_world_coordinates(center_xy, h_matrix) if scene == "0002": # 0002 scene is origin 1280x720, image is resized to 1920x1080 x, y = center_xy center_xy = [x * (1920 / 1280.0), y * (1920 / 1280.0)] world_xy = [-world_xy[0], world_xy[1]] # actev h is mirrored data_pixel.append([frame_index, track_id] + center_xy) data_world.append([frame_index, track_id] + world_xy) data_pixel.sort(key=operator.itemgetter(0)) data_world.sort(key=operator.itemgetter(0)) save_file(data_pixel, out_path_pixel, videoname) save_file(data_world, out_path_world, videoname)