读iphone的livelink输出bs的csv文件_csv live link face-CSDN博客

本文链接：https://blog.csdn.net/chenguowen21/article/details/131962018

该函数从CSV文件中读取数据，将时间转换为指定的帧率，然后使用线性插值对数据进行重新映射。它返回处理后的融合形状名称、重映射后的值和时间向量。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

def readblendfromcsv_fps(csvname, fps=60):
    file = open(csvname)
    csvreader = csv.reader(file)
    blend_name_src = next(csvreader)[2:]
    blend_shape_value_src = []
    times_vec = []
    for row in csvreader:
        if len(row[2:]) > 0:
            time_now = time_convert(row[0])
            times_vec.append(time_now)
            value = list(map(float, row[2:]))
            blend_shape_value_src.append(value)

    times_vec = np.array(times_vec) * fps
    times_vec = times_vec - times_vec[0]

    blend_shape_value_src = np.array(blend_shape_value_src)
    remap_len = int(times_vec[-1])
    remapped_blend = np.zeros((remap_len, len(blend_shape_value_src[0])))
    for i in range(remap_len):
        vect = times_vec < i
        # print(i, '#########################################  ', np.sum(vect))
        if np.sum(vect) == 0:
            id0 = 0
            id1 = 1
        else:
            id0 = np.sum(vect) - 1
            id1 = np.sum(vect)
        time0 = times_vec[id0]
        time1 = times_vec[id1]
        # print("time0=", time0, "   ", "time1=", time1)
        # print("i=", i)
        remapped_blend[i] = blend_shape_value_src[id0] * (time1 - i) + blend_shape_value_src[id1] * (i - time0)
        remapped_blend[i] = remapped_blend[i] / (time1 - time0)
    file.close()
    return blend_name_src, remapped_blend, times_vec