def print_bone_length_acc()

in src/eval.py [0:0]

• 45 lines of code
• 9 McCabe index (conditional complexity)

def print_bone_length_acc(bone_length_acc_ratio, bone_length_acc_distance, c: Train_Config, bone_lengths):
    body_acc_ratio = None
    body_acc_distance = None
    body_lengths = None
    all_ratio = None
    all_distance = None
    all_lengths = None

    joint_groups = get_joint_group_idx(c)
    group_bone_lengths = get_empty_joint_acc_dict(c)
    for group_name, joints in joint_groups.items():
        for j in joints:
            group_bone_lengths[group_name].append(bone_lengths[j].item())

    for k in bone_length_acc_ratio.keys():
        acc_ratio = torch.mean(torch.cat(bone_length_acc_ratio[k], dim=0)).item()
        rmse_acc_distance = torch.sqrt(torch.mean(torch.cat(bone_length_acc_distance[k], dim=0) ** 2)).item()
        mean_acc_distance = torch.mean(torch.abs(torch.cat(bone_length_acc_distance[k], dim=0))).item()
        avg_bone_length = torch.mean(torch.tensor(group_bone_lengths[k])).item()

        if all_ratio is None:
            all_ratio = torch.cat(bone_length_acc_ratio[k], dim=0)
            all_distance = torch.cat(bone_length_acc_distance[k], dim=0)
            all_lengths = torch.tensor(group_bone_lengths[k])
        else:
            all_ratio = torch.cat([all_ratio, *bone_length_acc_ratio[k]], dim=0)
            all_distance = torch.cat([all_distance, *bone_length_acc_distance[k]], dim=0)
            all_lengths = torch.cat([all_lengths, torch.tensor(group_bone_lengths[k])], dim=0)

        if not k.startswith("finger") and k != "thumb" and k != "wrist_local_finger":
            if body_acc_ratio is None:
                body_acc_ratio = torch.cat(bone_length_acc_ratio[k], dim=0)
                body_acc_distance = torch.cat(bone_length_acc_distance[k], dim=0)
                body_length = torch.tensor(group_bone_lengths[k])
            else:
                body_acc_ratio = torch.cat([body_acc_ratio, *bone_length_acc_ratio[k]], dim=0)
                body_acc_distance = torch.cat([body_acc_distance, *bone_length_acc_distance[k]], dim=0)
                body_length = torch.cat([body_length, torch.tensor(group_bone_lengths[k])], dim=0)

        print(f"{k:<20}mean ratio: {acc_ratio:.3f}   distance: {rmse_acc_distance:.4f}/{mean_acc_distance:.4f}   avg length: {avg_bone_length:.4f}")

    body_acc_ratio = torch.mean(body_acc_ratio).item()
    rmse_body_acc_distance = torch.sqrt(torch.mean(body_acc_distance ** 2)).item()
    mean_body_acc_distance = torch.mean(torch.abs(body_acc_distance)).item()
    body_length = torch.mean(body_length).item()
    all_acc_ratio = torch.mean(all_ratio).item()
    rmse_all_acc_distance = torch.sqrt(torch.mean(all_distance ** 2)).item()
    mean_all_acc_distance = torch.mean(torch.abs(all_distance)).item()
    all_length = torch.mean(all_lengths).item()
    print(f"{'body':<20}mean ratio: {body_acc_ratio:.3f}   distance: {rmse_body_acc_distance:.4f}/{mean_body_acc_distance:.4f}   avg length: {body_length:.4f}")
    print(f"{'all':<20}mean ratio: {all_acc_ratio:.3f}   distance: {rmse_all_acc_distance:.4f}/{mean_all_acc_distance:.4f}   avg length: {all_length:.4f}")