Loss Weights in repository different from those in your paper about deep3dfacerecon_pytorch

[sicxu]

You may adjust the loss weight for a quick check. 73d4911#r55395513

Deep3DFaceRecon_pytorch/models/facerecon_model.py

Lines 50 to 60 in 091e845

	# loss weights
	parser.add_argument('--w_feat', type=float, default=0.2, help='weight for feat loss')
	parser.add_argument('--w_color', type=float, default=1.92, help='weight for loss loss')
	parser.add_argument('--w_reg', type=float, default=3.0e-4, help='weight for reg loss')
	parser.add_argument('--w_id', type=float, default=1.0, help='weight for id_reg loss')
	parser.add_argument('--w_exp', type=float, default=0.8, help='weight for exp_reg loss')
	parser.add_argument('--w_tex', type=float, default=1.7e-2, help='weight for tex_reg loss')
	parser.add_argument('--w_gamma', type=float, default=10.0, help='weight for gamma loss')
	parser.add_argument('--w_lm', type=float, default=1.6e-3, help='weight for lm loss')
	parser.add_argument('--w_reflc', type=float, default=5.0, help='weight for reflc loss')

A large feature/perceptual loss may lead to bad performance. We tried to increase the w_feat, the reconstruction error increased on the MICC Florence dataset.

[humannumber196]

You may adjust the loss weight for a quick check. 73d4911#r55395513

Deep3DFaceRecon_pytorch/models/facerecon_model.py

Lines 50 to 60 in 091e845

	# loss weights
	parser.add_argument('--w_feat', type=float, default=0.2, help='weight for feat loss')
	parser.add_argument('--w_color', type=float, default=1.92, help='weight for loss loss')
	parser.add_argument('--w_reg', type=float, default=3.0e-4, help='weight for reg loss')
	parser.add_argument('--w_id', type=float, default=1.0, help='weight for id_reg loss')
	parser.add_argument('--w_exp', type=float, default=0.8, help='weight for exp_reg loss')
	parser.add_argument('--w_tex', type=float, default=1.7e-2, help='weight for tex_reg loss')
	parser.add_argument('--w_gamma', type=float, default=10.0, help='weight for gamma loss')
	parser.add_argument('--w_lm', type=float, default=1.6e-3, help='weight for lm loss')
	parser.add_argument('--w_reflc', type=float, default=5.0, help='weight for reflc loss')

A large feature/perceptual loss may lead to bad performance. We tried to increase the w_feat, the reconstruction error increased on the MICC Florence dataset.

OK, thanks for your reply!

[LY-Donkey-Burger]

You may adjust the loss weight for a quick check. 73d4911#r55395513

Deep3DFaceRecon_pytorch/models/facerecon_model.py

Lines 50 to 60 in 091e845

	# loss weights
	parser.add_argument('--w_feat', type=float, default=0.2, help='weight for feat loss')
	parser.add_argument('--w_color', type=float, default=1.92, help='weight for loss loss')
	parser.add_argument('--w_reg', type=float, default=3.0e-4, help='weight for reg loss')
	parser.add_argument('--w_id', type=float, default=1.0, help='weight for id_reg loss')
	parser.add_argument('--w_exp', type=float, default=0.8, help='weight for exp_reg loss')
	parser.add_argument('--w_tex', type=float, default=1.7e-2, help='weight for tex_reg loss')
	parser.add_argument('--w_gamma', type=float, default=10.0, help='weight for gamma loss')
	parser.add_argument('--w_lm', type=float, default=1.6e-3, help='weight for lm loss')
	parser.add_argument('--w_reflc', type=float, default=5.0, help='weight for reflc loss')

A large feature/perceptual loss may lead to bad performance. We tried to increase the w_feat, the reconstruction error increased on the MICC Florence dataset.

OK, thanks for your reply!

Hi ，could you get a model as good as the one published by the author ？

[humannumber196]

You may adjust the loss weight for a quick check. 73d4911#r55395513

Deep3DFaceRecon_pytorch/models/facerecon_model.py

Lines 50 to 60 in 091e845

	# loss weights
	parser.add_argument('--w_feat', type=float, default=0.2, help='weight for feat loss')
	parser.add_argument('--w_color', type=float, default=1.92, help='weight for loss loss')
	parser.add_argument('--w_reg', type=float, default=3.0e-4, help='weight for reg loss')
	parser.add_argument('--w_id', type=float, default=1.0, help='weight for id_reg loss')
	parser.add_argument('--w_exp', type=float, default=0.8, help='weight for exp_reg loss')
	parser.add_argument('--w_tex', type=float, default=1.7e-2, help='weight for tex_reg loss')
	parser.add_argument('--w_gamma', type=float, default=10.0, help='weight for gamma loss')
	parser.add_argument('--w_lm', type=float, default=1.6e-3, help='weight for lm loss')
	parser.add_argument('--w_reflc', type=float, default=5.0, help='weight for reflc loss')

A large feature/perceptual loss may lead to bad performance. We tried to increase the w_feat, the reconstruction error increased on the MICC Florence dataset.

OK, thanks for your reply!

Hi ，could you get a model as good as the one published by the author ？

I can't...

[LY-Donkey-Burger]

You may adjust the loss weight for a quick check. 73d4911#r55395513

Deep3DFaceRecon_pytorch/models/facerecon_model.py

Lines 50 to 60 in 091e845

	# loss weights
	parser.add_argument('--w_feat', type=float, default=0.2, help='weight for feat loss')
	parser.add_argument('--w_color', type=float, default=1.92, help='weight for loss loss')
	parser.add_argument('--w_reg', type=float, default=3.0e-4, help='weight for reg loss')
	parser.add_argument('--w_id', type=float, default=1.0, help='weight for id_reg loss')
	parser.add_argument('--w_exp', type=float, default=0.8, help='weight for exp_reg loss')
	parser.add_argument('--w_tex', type=float, default=1.7e-2, help='weight for tex_reg loss')
	parser.add_argument('--w_gamma', type=float, default=10.0, help='weight for gamma loss')
	parser.add_argument('--w_lm', type=float, default=1.6e-3, help='weight for lm loss')
	parser.add_argument('--w_reflc', type=float, default=5.0, help='weight for reflc loss')

A large feature/perceptual loss may lead to bad performance. We tried to increase the w_feat, the reconstruction error increased on the MICC Florence dataset.

OK, thanks for your reply!

Hi ，could you get a model as good as the one published by the author ？

I can't...

How much the reconstructed error of the model you get is different from the author's