def eval_proposed(weight, use_haptic, use_audio, use_virbo):
opt = Options().parse()
opt.use_behavior = True
opt.use_haptic = use_haptic
opt.use_audio = use_audio
opt.use_vibro = use_virbo
opt.aux = True
opt.sequence_length = 20
print("Model Config: ", opt)
model = Model(opt)
model.load_weight(weight)
return model.evaluate(0, keep_batch=True, ssim=False)
def eval_proposed(weight, use_haptic, use_audio, use_virbo, behavior):
opt = Options().parse()
opt.use_behavior = True
opt.use_haptic = use_haptic
opt.use_audio = use_audio
opt.use_vibro = False
opt.behavior_layer = 1
opt.aux = True
opt.sequence_length = 20
opt.data_dir = '../data/' + behavior
print("Model Config: ", opt)
model = Model(opt)
model.load_weight(weight)
return model.evaluate(0, keep_frame=True)
def predict_proposed(weight, use_haptic, use_audio, use_virbo, filelist):
opt = Options().parse()
opt.use_behavior = True
opt.use_haptic = use_haptic
opt.use_audio = use_audio
opt.use_vibro = use_virbo
opt.aux = True
opt.sequence_length = 20
print("Model Config: ", opt)
model = Model(opt)
model.load_weight(weight)
resultlist, gt = model.predict(filelist)
gen_audios = [np.vstack([hp.cpu().numpy().squeeze()[-3:] for hp in haptic]) for _, haptic, _, _ in resultlist]
gt_audios = [np.vstack([hp.cpu().numpy().squeeze()[:,-3:] for hp in haptic]) for _, haptic, _, _ in gt]
pass
