Method load_weights_from_keras

script/models/nerf.py:241–268

(self, weights)

        return outputs

    def load_weights_from_keras(self, weights):
        assert self.use_viewdirs, "Not implemented if use_viewdirs=False"

        # Load pts_linears
        for i in range(self.D):
            idx_pts_linears = 2 * i
            self.pts_linears[i].weight.data = torch.from_numpy(np.transpose(weights[idx_pts_linears]))
            self.pts_linears[i].bias.data = torch.from_numpy(np.transpose(weights[idx_pts_linears+1]))

        # Load feature_linear
        idx_feature_linear = 2 * self.D
        self.feature_linear.weight.data = torch.from_numpy(np.transpose(weights[idx_feature_linear]))
        self.feature_linear.bias.data = torch.from_numpy(np.transpose(weights[idx_feature_linear+1]))

        # Load views_linears
        idx_views_linears = 2 * self.D + 2
        self.views_linears[0].weight.data = torch.from_numpy(np.transpose(weights[idx_views_linears]))
        self.views_linears[0].bias.data = torch.from_numpy(np.transpose(weights[idx_views_linears+1]))

        # Load rgb_linear
        idx_rbg_linear = 2 * self.D + 4
        self.rgb_linear.weight.data = torch.from_numpy(np.transpose(weights[idx_rbg_linear]))
        self.rgb_linear.bias.data = torch.from_numpy(np.transpose(weights[idx_rbg_linear+1]))

        # Load alpha_linear
        idx_alpha_linear = 2 * self.D + 6
        self.alpha_linear.weight.data = torch.from_numpy(np.transpose(weights[idx_alpha_linear]))
        self.alpha_linear.bias.data = torch.from_numpy(np.transpose(weights[idx_alpha_linear+1]))

def create_nerf(args):
    """Instantiate NeRF's MLP model.