reparo.order¶
-
reparo.order.chain_frames(dist_matrix, first_frame=0, valid_idxs=None)[source]¶ Chain together frames that are similar
- Parameters
dist_matrix (numpy.ndarray[N, N]) – frame codistance matrix
first_frame (int, optional) – index of the first frame
valid_idxs (list<int>, optional) – list of valid frame indices
- Returns
ordered frame indices
- Return type
list<int>
-
reparo.order.get_chain_split(used_frames, dist_matrix)[source]¶ Filter noisy frames out
- Parameters
used_frames (list<int>) – ordered frame indices
dist_matrix (numpy.ndarray[N, N]) – frame codistance matrix
- Returns
pair of frame indices around the sequence split
- Return type
tuple<int, int>
-
reparo.order.reorder_frames(frames, metric='cosine', target_size=(128, 128), denoise=False)[source]¶ Reorder frames in a coherent fashion
- Parameters
frames (list<numpy.ndarray[H, W, 3]>) – video frames
metric (str, optional) – matrix used for codistance matrix
target_size (tuple<int, int>, optional) – frame size used for codistance computation
denoise (bool, optional) – whether the frames need to be denoised
- Returns
video frames
- Return type
list<numpy.ndarray[H, W, 3]>