model now uses tensorflow dataset generator
This commit is contained in:
parent
ba6c132c67
commit
f06d3ae504
2 changed files with 84 additions and 48 deletions
|
|
@ -27,7 +27,7 @@ if gpus:
|
|||
print(e)
|
||||
|
||||
|
||||
from video_compression_model import VideoCompressionModel, data_generator
|
||||
from video_compression_model import VideoCompressionModel, create_dataset
|
||||
|
||||
from globalVars import HEIGHT, WIDTH, MAX_FRAMES, LOGGER
|
||||
|
||||
|
|
@ -43,7 +43,7 @@ EARLY_STOP = 5
|
|||
|
||||
class GarbageCollectorCallback(Callback):
|
||||
def on_epoch_end(self, epoch, logs=None):
|
||||
LOGGER.debug(f"Collecting garbage")
|
||||
LOGGER.debug(f"GC")
|
||||
gc.collect()
|
||||
|
||||
def save_model(model):
|
||||
|
|
@ -120,6 +120,10 @@ def main():
|
|||
split_index = int(0.8 * len(all_videos))
|
||||
training_videos = all_videos[:split_index]
|
||||
validation_videos = all_videos[split_index:]
|
||||
|
||||
training_dataset = create_dataset(training_videos, BATCH_SIZE, MAX_FRAMES)
|
||||
validation_dataset = create_dataset(validation_videos, BATCH_SIZE, MAX_FRAMES)
|
||||
|
||||
|
||||
if args.continue_training:
|
||||
model = tf.keras.models.load_model(args.continue_training)
|
||||
|
|
@ -154,26 +158,24 @@ def main():
|
|||
gc_callback = GarbageCollectorCallback()
|
||||
|
||||
# Calculate steps per epoch for training and validation
|
||||
if MAX_FRAMES <= 0:
|
||||
average_frames_per_video = 2880 # Given 2 minutes @ 24 fps
|
||||
else:
|
||||
average_frames_per_video = max(MAX_FRAMES, 0)
|
||||
#if MAX_FRAMES <= 0:
|
||||
# average_frames_per_video = 2880 # Given 2 minutes @ 24 fps
|
||||
#else:
|
||||
# average_frames_per_video = max(MAX_FRAMES, 0)
|
||||
|
||||
total_frames_train = average_frames_per_video * len(training_videos)
|
||||
total_frames_validation = average_frames_per_video * len(validation_videos)
|
||||
steps_per_epoch_train = total_frames_train // BATCH_SIZE
|
||||
steps_per_epoch_validation = total_frames_validation // BATCH_SIZE
|
||||
#total_frames_train = average_frames_per_video * len(training_videos)
|
||||
#total_frames_validation = average_frames_per_video * len(validation_videos)
|
||||
#steps_per_epoch_train = total_frames_train // BATCH_SIZE
|
||||
#steps_per_epoch_validation = total_frames_validation // BATCH_SIZE
|
||||
|
||||
gc.collect()
|
||||
|
||||
# Train the model
|
||||
LOGGER.info("Starting model training.")
|
||||
model.fit(
|
||||
data_generator(training_videos, BATCH_SIZE),
|
||||
epochs=EPOCHS,
|
||||
steps_per_epoch=steps_per_epoch_train,
|
||||
validation_data=data_generator(validation_videos, BATCH_SIZE), # Add validation data here
|
||||
validation_steps=steps_per_epoch_validation, # Add validation steps here
|
||||
training_dataset,
|
||||
epochs=EPOCHS,
|
||||
validation_data=validation_dataset, # Add validation data here
|
||||
callbacks=[early_stop, checkpoint_callback, gc_callback]
|
||||
)
|
||||
LOGGER.info("Model training completed.")
|
||||
|
|
|
|||
|
|
@ -28,52 +28,86 @@ def combine_batch(frame, crf, speed, include_controls=True, resize=True):
|
|||
return np.concatenate(combined, axis=-1)
|
||||
|
||||
|
||||
def data_generator(videos, batch_size):
|
||||
def process_video(video):
|
||||
base_dir = os.path.dirname("test_data/validation/validation.json")
|
||||
|
||||
cap_compressed = cv2.VideoCapture(os.path.join(base_dir, video["compressed_video_file"]))
|
||||
cap_uncompressed = cv2.VideoCapture(os.path.join(base_dir, video["original_video_file"]))
|
||||
|
||||
compressed_frames = []
|
||||
uncompressed_frames = []
|
||||
|
||||
while True:
|
||||
# Lists to store the processed frames
|
||||
compressed_frame_batch = [] # Input data (Target)
|
||||
uncompressed_frame_batch = [] # Target data (Training)
|
||||
ret_compressed, compressed_frame = cap_compressed.read()
|
||||
ret_uncompressed, uncompressed_frame = cap_uncompressed.read()
|
||||
|
||||
# Get a list of video capture objects for all videos
|
||||
caps_compressed = [cv2.VideoCapture(os.path.join(base_dir, video["compressed_video_file"])) for video in videos]
|
||||
caps_uncompressed = [cv2.VideoCapture(os.path.join(base_dir, video["original_video_file"])) for video in videos]
|
||||
if not ret_compressed or not ret_uncompressed:
|
||||
break
|
||||
|
||||
# As long as any video can provide frames, keep running
|
||||
while any(cap.isOpened() for cap in caps_compressed):
|
||||
for idx, (cap_compressed, cap_uncompressed) in enumerate(zip(caps_compressed, caps_uncompressed)):
|
||||
#print(f"(Video Change) Processing video {idx}") # Print statement to indicate video change
|
||||
|
||||
ret_compressed, compressed_frame = cap_compressed.read()
|
||||
ret_uncompressed, uncompressed_frame = cap_uncompressed.read()
|
||||
CRF = scale_crf(video["crf"])
|
||||
SPEED = scale_speed_preset(PRESET_SPEED_CATEGORIES.index(video["preset_speed"]))
|
||||
|
||||
if not ret_compressed or not ret_uncompressed:
|
||||
continue
|
||||
compressed_combined = combine_batch(compressed_frame, CRF, SPEED, include_controls=False)
|
||||
uncompressed_combined = combine_batch(uncompressed_frame, 0, scale_speed_preset(PRESET_SPEED_CATEGORIES.index("veryslow")))
|
||||
|
||||
CRF = scale_crf(videos[idx]["crf"])
|
||||
SPEED = scale_speed_preset(PRESET_SPEED_CATEGORIES.index(videos[idx]["preset_speed"]))
|
||||
compressed_frames.append(compressed_combined)
|
||||
uncompressed_frames.append(uncompressed_combined)
|
||||
|
||||
compressed_combined = combine_batch(compressed_frame, CRF, SPEED, include_controls=False)
|
||||
uncompressed_combined = combine_batch(uncompressed_frame, 0, scale_speed_preset(PRESET_SPEED_CATEGORIES.index("veryslow")))
|
||||
cap_compressed.release()
|
||||
cap_uncompressed.release()
|
||||
|
||||
compressed_frame_batch.append(compressed_combined)
|
||||
uncompressed_frame_batch.append(uncompressed_combined)
|
||||
return uncompressed_frames, compressed_frames
|
||||
|
||||
if len(compressed_frame_batch) == batch_size:
|
||||
yield (np.array(uncompressed_frame_batch), np.array(compressed_frame_batch))
|
||||
compressed_frame_batch.clear()
|
||||
uncompressed_frame_batch.clear()
|
||||
|
||||
# Close all video captures at the end
|
||||
for cap in caps_compressed + caps_uncompressed:
|
||||
cap.release()
|
||||
def frame_generator(videos, max_frames=None):
|
||||
base_dir = "test_data/validation/"
|
||||
for video in videos:
|
||||
cap_compressed = cv2.VideoCapture(os.path.join(base_dir, video["compressed_video_file"]))
|
||||
cap_uncompressed = cv2.VideoCapture(os.path.join(base_dir, video["original_video_file"]))
|
||||
|
||||
cv2.destroyAllWindows()
|
||||
frame_count = 0
|
||||
while True:
|
||||
ret_compressed, compressed_frame = cap_compressed.read()
|
||||
ret_uncompressed, uncompressed_frame = cap_uncompressed.read()
|
||||
|
||||
if not ret_compressed or not ret_uncompressed:
|
||||
break
|
||||
|
||||
CRF = scale_crf(video["crf"])
|
||||
SPEED = scale_speed_preset(PRESET_SPEED_CATEGORIES.index(video["preset_speed"]))
|
||||
|
||||
compressed_combined = combine_batch(compressed_frame, CRF, SPEED, include_controls=False)
|
||||
uncompressed_combined = combine_batch(uncompressed_frame, 0, scale_speed_preset(PRESET_SPEED_CATEGORIES.index("veryslow")))
|
||||
|
||||
yield uncompressed_combined, compressed_combined
|
||||
|
||||
frame_count += 1
|
||||
if max_frames is not None and frame_count >= max_frames:
|
||||
break
|
||||
|
||||
cap_compressed.release()
|
||||
cap_uncompressed.release()
|
||||
|
||||
|
||||
|
||||
def create_dataset(videos, batch_size, max_frames=None):
|
||||
# Determine the output signature by processing a single video to obtain its shape
|
||||
video_generator_instance = frame_generator(videos, max_frames)
|
||||
sample_uncompressed, sample_compressed = next(video_generator_instance)
|
||||
output_signature = (
|
||||
tf.TensorSpec(shape=tf.shape(sample_uncompressed), dtype=tf.float32),
|
||||
tf.TensorSpec(shape=tf.shape(sample_compressed), dtype=tf.float32)
|
||||
)
|
||||
|
||||
dataset = tf.data.Dataset.from_generator(
|
||||
lambda: frame_generator(videos, max_frames), # Include max_frames argument through lambda
|
||||
output_signature=output_signature
|
||||
)
|
||||
|
||||
dataset = dataset.shuffle(100).batch(batch_size).prefetch(tf.data.experimental.AUTOTUNE)
|
||||
|
||||
return dataset
|
||||
|
||||
# If there are frames left that don't fill a whole batch, send them anyway
|
||||
if len(compressed_frame_batch) > 0:
|
||||
yield (np.array(uncompressed_frame_batch), np.array(compressed_frame_batch))
|
||||
|
||||
|
||||
class VideoCompressionModel(tf.keras.Model):
|
||||
|
|
|
|||
Reference in a new issue