import mmpose
import os
from mmpose.apis import MMPoseInferencer
print("[INFO]: Imported modules!")

import gradio as gr
import numpy as np
import cv2

# inferencer = MMPoseInferencer('hand') # 'hand', 'human , device='cuda'
# inferencer = MMPoseInferencer('human')

inferencer = MMPoseInferencer(pose3d='human3d')
# https://github.com/open-mmlab/mmpose/tree/dev-1.x/configs/body_3d_keypoint/pose_lift
# motionbert_ft_h36m-d80af323_20230531.pth
# simple3Dbaseline_h36m-f0ad73a4_20210419.pth
# videopose_h36m_243frames_fullconv_supervised_cpn_ft-88f5abbb_20210527.pth
# videopose_h36m_81frames_fullconv_supervised-1f2d1104_20210527.pth
# videopose_h36m_27frames_fullconv_supervised-fe8fbba9_20210527.pth
# videopose_h36m_1frame_fullconv_supervised_cpn_ft-5c3afaed_20210527.pth

# https://github.com/open-mmlab/mmpose/blob/main/mmpose/apis/inferencers/pose3d_inferencer.py

print("[INFO]: Downloaded models!")


def poses(photo):
    print(photo)
    result_generator = inferencer(photo, 
                                  vis_out_dir =".",
                                  return_vis=True,
                                  thickness=2)
    
    

    # Prepare to save video
    output_file = os.path.join("output.mp4")

    fourcc = cv2.VideoWriter_fourcc(*"mp4v")  # Codec for MP4 video
    fps = 32
    height = 480
    width = 640
    size = (width,height)

    out_writer = cv2.VideoWriter(output_file, fourcc, fps, size)

    for result in result_generator:
        print("[INFO] Result: ", result)
        frame = result["visualization"]
        out_writer.write(cv2.cvtColor(frame[0], cv2.COLOR_BGR2RGB))

    print(os.listdir())
    print("[INFO]: Visualizing results!")
    print(os.listdir())
    print()

    out_writer.release()
    cv2.destroyAllWindows() # Closing window


    return output_file


# # specify detection model by alias
# # the available aliases include 'human', 'hand', 'face', 'animal',
# # as well as any additional aliases defined in mmdet
# inferencer = MMPoseInferencer(
#     # suppose the pose estimator is trained on custom dataset
#     pose2d='custom_human_pose_estimator.py',
#     pose2d_weights='custom_human_pose_estimator.pth',
#     det_model='human'
# )
    
def run():
    #https://github.com/open-mmlab/mmpose/blob/main/docs/en/user_guides/inference.md
    demo = gr.Interface(fn=poses, 
                        inputs=gr.Video(source="upload"),
                        outputs=gr.Video())


    demo.launch(server_name="0.0.0.0", server_port=7860)


if __name__ == "__main__":
    run()
    print(os.listdir())