def cross_validate(self, event):
trainingsetindex = self.trainingset.GetValue()
shuffle = [self.shuffles.GetValue()]
cfg = auxiliaryfunctions.read_config(self.config)
trainFraction = cfg["TrainingFraction"][trainingsetindex]
self.inf_cfg_path = os.path.join(
cfg["project_path"],
auxiliaryfunctions.GetModelFolder(trainFraction,
self.shuffles.GetValue(), cfg),
"test",
"inference_cfg.yaml",
)
# Read from edited inf. file first ...
print(self.inf_cfg_path)
print("optimizing parameters using " + self.targettypes.GetValue() +
" as a target...")
deeplabcut.evaluate_multianimal_crossvalidate(
self.config,
Shuffles=shuffle,
trainingsetindex=trainingsetindex,
edgewisecondition=self.edgeWise.GetStringSelection(),
leastbpts=self.infg.GetValue(),
init_points=self.inpts.GetValue(),
n_iter=self.n_iter.GetValue(),
target=self.targettypes.GetValue(),
)
"global_scale": 0.5,
"batch_size": 1,
"save_iters": N_ITER,
"display_iters": N_ITER // 2,
# "multi_step": [[0.001, N_ITER]],
}
deeplabcut.auxiliaryfunctions.edit_config(pose_config_path, edits)
print("Pose config edited.")
print("Training network...")
deeplabcut.train_network(config_path, maxiters=N_ITER)
print("Network trained.")
print("Evaluating network...")
deeplabcut.evaluate_network(config_path, plotting=True)
deeplabcut.evaluate_multianimal_crossvalidate(
config_path, n_iter=8, init_points=3) # parameters so it is fast
print("Network evaluated....")
new_video_path = deeplabcut.ShortenVideo(
video_path,
start="00:00:00",
stop="00:00:01",
outsuffix="short",
outpath=os.path.join(cfg["project_path"], "videos"),
)
print("Analyzing video...")
deeplabcut.analyze_videos(config_path, [new_video_path],
"mp4",
robust_nframes=True)