NSAQA / tempSegAndAllErrorsForAllFrames_newVids.py
laurenok24's picture
Update tempSegAndAllErrorsForAllFrames_newVids.py
ba67987 verified
raw
history blame
15.2 kB
from microprograms.temporal_segmentation.entry import entry_microprogram_one_frame
from microprograms.temporal_segmentation.somersault import somersault_microprogram_one_frame
from microprograms.temporal_segmentation.twist import twist_microprogram_one_frame
from microprograms.temporal_segmentation.start_takeoff import takeoff_microprogram_one_frame
from microprograms.errors.distance_from_springboard_micro_program import board_end
from microprograms.errors.splash_micro_program import *
from microprograms.errors.distance_from_springboard_micro_program import calculate_distance_from_springboard_for_one_frame
from microprograms.errors.distance_from_springboard_micro_program import calculate_distance_from_platform_for_one_frame
from microprograms.errors.distance_from_springboard_micro_program import find_which_side_board_on
from microprograms.errors.angles_micro_programs import applyFeetApartError
from microprograms.errors.angles_micro_programs import applyPositionTightnessError
from models.detectron2.platform_detector_setup import get_platform_detector
from models.pose_estimator.pose_estimator_model_setup import get_pose_estimation
from models.detectron2.diver_detector_setup import get_diver_detector
from models.pose_estimator.pose_estimator_model_setup import get_pose_model
from models.detectron2.splash_detector_setup import get_splash_detector
from somersault_counter import som_counter, twist_counter
from microprograms.errors.over_rotation import over_rotation
from temporal_segmentation import detect_on_board
from dive_recognition_functions import *
from scoring_functions import get_scale_factor
import gradio as gr
import pickle
import os
import math
import numpy as np
import cv2
# with open('segmentation_error_data.pkl', 'rb') as f:
# data = pickle.load(f)
def getDiveInfo_from_diveNum(diveNum):
handstand = (diveNum[0] == '6')
expected_som = int(diveNum[2])
if len(diveNum) == 5:
expected_twists = int(diveNum[3])
else:
expected_twists = 0
if diveNum[0] == '1' or diveNum[0] == '3' or diveNum[:2] == '51' or diveNum[:2] == '53' or diveNum[:2] == '61' or diveNum[:2] == '63':
back_facing = False
else:
back_facing = True
if diveNum[0] == '1' or diveNum[:2] == '51' or diveNum[:2] == '61':
expected_direction = 'front'
elif diveNum[0] == '2' or diveNum[:2] == '52' or diveNum[:2] == '62':
expected_direction = 'back'
elif diveNum[0] == '3' or diveNum[:2] == '53' or diveNum[:2] == '63':
expected_direction = 'reverse'
elif diveNum[0] == '4':
expected_direction = 'inward'
if diveNum[-1] == 'b':
position = 'pike'
elif diveNum[-1] == 'c':
position = 'tuck'
else:
position = 'free'
return handstand, expected_som, expected_twists, back_facing, expected_direction, position
def getDiveInfo_from_symbols(frames, dive_data=None, platform_detector=None, splash_detector=None, diver_detector=None, pose_model=None):
print("Getting dive info from symbols...")
if dive_data is None:
print("somethings not getting passed in properly")
dive_data = abstractSymbols(frames, platform_detector=platform_detector, splash_detector=splash_detector, diver_detector=diver_detector, pose_model=pose_model)
# get above_boards, on_boards, and position_tightness
above_board = True
on_board = True
above_boards = []
on_boards = []
position_tightness = []
distances = []
prev_board_coord = None
for i in range(len(dive_data['pose_pred'])):
pose_pred = dive_data['pose_pred'][i]
board_end_coord = dive_data['board_end_coords'][i]
if board_end_coord is not None and prev_board_coord is not None:
distances.append(math.dist(board_end_coord, prev_board_coord))
if math.dist(board_end_coord, prev_board_coord) > 150:
position_tightness.append(applyPositionTightnessError(filepath="", pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model))
if above_board:
above_boards.append(1)
else:
above_boards.append(0)
if on_board:
on_boards.append(1)
else:
on_boards.append(0)
continue
if above_board and not on_board and board_end_coord is not None and pose_pred is not None and np.array(pose_pred)[0][2][1] > int(board_end_coord[1]):
above_board=False
if on_board:
handstand = is_handstand(dive_data)
calculate_on_board = detect_on_board(board_end_coord, dive_data['board_side'], pose_pred, handstand)
if calculate_on_board is not None and not calculate_on_board:
on_board = False
if above_board:
above_boards.append(1)
else:
above_boards.append(0)
if on_board:
on_boards.append(1)
else:
on_boards.append(0)
prev_board_coord = board_end_coord
position_tightness.append(applyPositionTightnessError(filepath="", pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model))
dive_data['on_boards'] = on_boards
dive_data['above_boards'] = above_boards
dive_data['position_tightness'] = position_tightness
## handstand and som_count##
expected_som, handstand = som_counter_full_dive(dive_data)
## twist_count
expected_twists = twist_counter_full_dive(dive_data)
## direction: front, back, reverse, inward
expected_direction = get_direction(dive_data)
return handstand, expected_som, expected_twists, expected_direction, dive_data
def abstractSymbols(frames, progress=gr.Progress(), platform_detector=None, splash_detector=None, diver_detector=None, pose_model=None):
print("Abstracting symbols...")
splashes = []
pose_preds = []
board_sides = []
plat_outputs = []
diver_boxes = []
splash_pred_masks = []
if platform_detector is None:
platform_detector = get_platform_detector()
if splash_detector is None:
splash_detector = get_splash_detector()
if diver_detector is None:
diver_detector = get_diver_detector()
if pose_model is None:
pose_model = get_pose_model()
num_frames = len(frames)
i = 0
for frame in frames:
progress(i/num_frames, desc="Abstracting Symbols")
plat_output = platform_detector(frame)
plat_outputs.append(plat_output)
board_side = find_which_side_board_on(plat_output)
if board_side is not None:
board_sides.append(board_side)
diver_box, pose_pred = get_pose_estimation(filepath="", image_bgr=frame, diver_detector=diver_detector, pose_model=pose_model)
pose_preds.append(pose_pred)
diver_boxes.append(diver_box)
splash_area, splash_pred_mask = get_splash_from_one_frame(filepath="", im=frame, predictor=splash_detector, visualize=False)
splash_pred_masks.append(splash_pred_mask)
splashes.append(splash_area)
i+=1
dive_data = {}
dive_data['plat_outputs'] = plat_outputs
dive_data['pose_pred'] = pose_preds
dive_data['splash'] = splashes
dive_data['splash_pred_masks'] = splash_pred_masks
dive_data['board_sides'] = board_sides
board_sides.sort()
board_side = board_sides[len(board_sides)//2]
dive_data['board_side'] = board_side
dive_data['diver_boxes'] = diver_boxes
# get board_end_coords
board_end_coords = []
for plat_output in dive_data['plat_outputs']:
board_end_coord = board_end(plat_output, board_side=dive_data['board_side'])
board_end_coords.append(board_end_coord)
dive_data['board_end_coords'] = board_end_coords
return dive_data
def getAllErrorsAndSegmentation_newVids(frames, dive_data, progress=gr.Progress(), diveNum="", board_side=None, platform_detector=None, splash_detector=None, diver_detector=None, pose_model=None):
print("in getAllErrorsAndSegmentation function...")
if len(frames) != len(dive_data['pose_pred']):
raise gr.Error("Abstract Symbols first!")
if diveNum != "":
dive_num_given = True
handstand, expected_som, expected_twists, back_facing, expected_direction, position = getDiveInfo_from_diveNum(diveNum)
else:
dive_num_given = False
handstand, expected_som, expected_twists, expected_direction, dive_data = getDiveInfo_from_symbols(frames, dive_data=dive_data, platform_detector=platform_detector, splash_detector=splash_detector, diver_detector=diver_detector, pose_model=pose_model)
if not dive_num_given:
above_boards = dive_data['above_boards']
on_boards = dive_data['on_boards']
position_tightness = dive_data['position_tightness']
board_end_coords = dive_data['board_end_coords']
else:
above_board = True
on_board = True
above_boards = []
on_boards = []
board_end_coords = []
position_tightness = []
splash = dive_data['splash']
diver_boxes = dive_data['diver_boxes']
board_side = dive_data['board_side']
pose_preds = dive_data['pose_pred']
takeoff = []
twist = []
som = []
entry = []
distance_from_board = []
feet_apart = []
over_under_rotation = []
som_counts = []
twist_counts = []
if platform_detector is None:
platform_detector = get_platform_detector()
if splash_detector is None:
splash_detector = get_splash_detector()
if diver_detector is None:
diver_detector = get_diver_detector()
if pose_model is None:
pose_model = get_pose_model()
j = 0
prev_pred = None
som_prev_pred = None
half_som_count=0
petal_count = 0
in_petal = False
num_frames = len(frames)
for i in range(num_frames):
progress(i/num_frames, desc="Calculating Dive Errors")
pose_pred = pose_preds[i]
calculated_half_som_count, skip = som_counter(pose_pred, prev_pose_pred=som_prev_pred, half_som_count=half_som_count, handstand=handstand)
if not skip:
som_prev_pred = pose_pred
calculated_petal_count, calculated_in_petal = twist_counter(pose_pred, prev_pose_pred=prev_pred, in_petal=in_petal, petal_count=petal_count)
if dive_num_given:
outputs = platform_detector(frames[i])
board_end_coord = board_end(outputs, board_side=board_side)
board_end_coords.append(board_end_coord)
if above_board and not on_board and board_end_coord is not None and pose_pred is not None and np.array(pose_pred)[0][2][1] > int(board_end_coord[1]):
above_board=False
if on_board and detect_on_board(board_end_coord, board_side, pose_pred, handstand) is not None and not detect_on_board(board_end_coord, board_side, pose_pred, handstand):
on_board = False
if above_board:
above_boards.append(1)
else:
above_boards.append(0)
if on_board:
on_boards.append(1)
else:
on_boards.append(0)
else:
board_end_coord = board_end_coords[i]
above_board = (above_boards[i] == 1)
on_board = (on_boards[i] == 1)
calculated_takeoff = takeoff_microprogram_one_frame(filepath="", above_board=above_board, on_board=on_board, pose_pred=pose_pred)
calculated_twist = twist_microprogram_one_frame(filepath="", on_board=on_board, pose_pred=pose_pred, expected_twists=expected_twists, petal_count=petal_count, expected_som=expected_som, half_som_count=half_som_count, diver_detector=diver_detector, pose_model=pose_model)
calculated_som = somersault_microprogram_one_frame(filepath="", pose_pred=pose_pred, on_board=on_board, expected_som=expected_som, half_som_count=half_som_count, expected_twists=expected_twists, petal_count=petal_count, diver_detector=diver_detector, pose_model=pose_model)
calculated_entry = entry_microprogram_one_frame(filepath="", frame=frames[i], above_board=above_board, on_board=on_board, pose_pred=pose_pred, expected_twists=expected_twists, petal_count=petal_count, expected_som=expected_som, half_som_count=half_som_count, splash_detector=splash_detector, visualize=False)
if calculated_som == 1:
half_som_count = calculated_half_som_count
elif calculated_twist == 1:
half_som_count = calculated_half_som_count
petal_count = calculated_petal_count
in_petal = calculated_in_petal
# distance from board
dist = calculate_distance_from_platform_for_one_frame(filepath="", im=frames[i], visualize=False, pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model, board_end_coord=board_end_coord, platform_detector=platform_detector) # saves photo to ./output/data/distance_from_board/
distance_from_board.append(dist)
if dive_num_given:
position_tightness.append(applyPositionTightnessError(filepath="", pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model))
# splash.append(get_splash_from_one_frame(filepath="", im=frames[i], predictor=splash_detector, visualize=False))
feet_apart.append(applyFeetApartError(filepath="", pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model))
over_under_rotation.append(over_rotation(filepath="", pose_pred=pose_pred, diver_detector=diver_detector, pose_model=pose_model))
takeoff.append(calculated_takeoff)
twist.append(calculated_twist)
som.append(calculated_som)
entry.append(calculated_entry)
som_counts.append(half_som_count)
twist_counts.append(petal_count)
prev_pred = pose_pred
print("takeoff", takeoff)
print("twist", twist)
print("som", som)
print("entry", entry)
print("distance_from_board", distance_from_board)
print("position_tightness", position_tightness)
print("feet_apart", feet_apart)
print("over_under_rotation", over_under_rotation)
print("splash", splash)
print("above_boards", above_boards)
print("on_boards", on_boards)
print("som_counts", som_counts)
print("twist_counts", twist_counts)
print("board_end_coords", board_end_coords)
print("diver_boxes", diver_boxes)
print("saving data into dive_data dictionary...")
dive_data['takeoff'] = takeoff
dive_data['twist'] = twist
dive_data['som'] = som
dive_data['entry'] = entry
dive_data['distance_from_board'] = distance_from_board
dive_data['position_tightness'] = position_tightness
dive_data['feet_apart'] = feet_apart
dive_data['over_under_rotation'] = over_under_rotation
dive_data['above_boards'] = above_boards
dive_data['on_boards'] = on_boards
dive_data['som_counts'] = som_counts
dive_data['twist_counts'] = twist_counts
dive_data['board_end_coords'] = board_end_coords
dive_data['is_handstand'] = handstand
dive_data['direction'] = expected_direction
return dive_data