首先這邊要去解讀這份MotionNet中資料前處理的gen_data.py程式碼

接下來會記錄每個區塊的函數以及變數所代表的意義

hyperparameters

scenes = np.load('data/split.npy', allow_pickle=True).item().get(args.split)
print("Split: {}, which contains {} scenes.".format(args.split, len(scenes)))

這邊讀取已經存下來的場景編號'scene_37' 'scene_34' 'scene_817'

gen_data()

    res_scenes = list()
    for s in scenes:
        s_id = s.split('_')[1]
        res_scenes.append(int(s_id))

把場景的編號存起來

for scene_idx in res_scenes:
        curr_scene = nusc.scene[scene_idx] #現在場景編號
        
        first_sample_token = curr_scene['first_sample_token'] #取出第一個sample token
        curr_sample = nusc.get('sample', first_sample_token) # 取出sample的資料
        curr_sample_data = nusc.get('sample_data', curr_sample['data']['LIDAR_TOP']) #取出sample中 lidar的資料

        save_data_dict_list = list()  # for storing consecutive sequences; the data consists of timestamps, points, etc
        save_box_dict_list = list()  # for storing box annotations in consecutive sequences
        save_instance_token_list = list()
        adj_seq_cnt = 0
        save_seq_cnt = 0  # only used for save data file name

 while curr_sample_data['next'] != '':

        # Get the synchronized point clouds
        all_pc, all_times, trans_matrices = \
        LidarPointCloud.from_file_multisweep_bf_sample_data
            (nusc, curr_sample_data,return_trans_matrix=True,                                   nsweeps_back=nsweeps_back,nsweeps_forward=nsweeps_forward)  
        
        # Store point cloud of each sweep
        pc = all_pc.points
        _, sort_idx = np.unique(all_times, return_index=True)
        unique_times = all_times[np.sort(sort_idx)]  # Preserve the item order in unique_times
        num_sweeps = len(unique_times)
        
        # 如果沒有足夠的past 跟 future sweeps的話 要跳過一些keyframe 
        #===================Skipping keyframe==============================
        # Make sure we have sufficient past and future sweeps
        if num_sweeps != (nsweeps_back + nsweeps_forward):

            # Skip some keyframes if necessary
            flag = False
            for _ in range(num_keyframe_skipped + 1):
                if curr_sample['next'] != '':
                    curr_sample = nusc.get('sample', curr_sample['next'])
                else:
                    flag = True
                    break

            if flag:  # No more keyframes
                break
            else:
                curr_sample_data = nusc.get('sample_data', curr_sample['data']['LIDAR_TOP'])

            # Reset
            adj_seq_cnt = 0
            save_data_dict_list = list()
            save_box_dict_list = list()
            save_instance_token_list = list()
            continue
      #===================Skipping keyframe==============================

LidarPointCloud.from_file_multisweep_bf_sample_data :
回傳aggregate完的pointcloud 簡單來說就是把過去的跟未來幾個frame的pointcloud同步到當下的frame 回傳的是很多point cloud組合而成的點雲集合 可以利用unique time來把它還原成對應時間的sweep

# Prepare data dictionary for the next step (ie, generating BEV maps)
            save_data_dict = dict()
            box_data_dict = dict()  # for remapping the instance ids, according to class_map
            curr_token_list = list()
                
            for tid in range(num_sweeps):
                _time = unique_times[tid]
                # 回傳相同元素的index位置
                points_idx = np.where(all_times == _time)[0]
                _pc = pc[:, points_idx] # 取出對應sweep時間當下的點雲(4,nbr_points)
                save_data_dict['pc_' + str(tid)] = _pc
            save_data_dict['times'] = unique_times
            save_data_dict['num_sweeps'] = num_sweeps
            save_data_dict['trans_matrices'] = trans_matrices
            

            # Get the synchronized bounding boxes
            # First, we need to iterate all the instances, and then retrieve their corresponding bounding boxes
            num_instances = 0  # The number of instances within this sample
            corresponding_sample_token = curr_sample_data['sample_token']
            corresponding_sample_rec = nusc.get('sample', corresponding_sample_token)
            
            

• 這邊的unique time 是指每個sweep對應到的時間 (長度=backward+forward sweeps nums)
• all_times是把每個sweep所有的點都對應到該sweep的時間 (ex:[1,1,1,1,1...1],總共23456個點就有23456個時間)
• 每個sample 中有每個sensor的資訊以及這個sample下的annotations以及sample當下的時間,及前跟後一個frame的token id
• print(save_data_dict.keys())會有 (pc_0 - pc_44 for test split), times, num_sweeps, trans_matrices