[Football Analysis] 14.위치 추적 데이터 가공 및 시각화

작성일자 : 2024-05-07

Ver 0.1.1

강의에서 소개된 파이썬 주요 기능¶

• ffmpeg: https://anaconda.org/conda-forge/ffmpeg
  • 터미널에서 conda install -c conda-forge ffmpeg 명령어 실행
• scipy.signal.savgol_filter: https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.savgol_filter.html
• matplotlib.pyplot.quiver: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.quiver.html
• matplotlib.animation.FFMpegWriter: https://matplotlib.org/stable/api/_as_gen/matplotlib.animation.FFMpegWriter.html
• matplotlib.pyplot.clf: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.clf.html
• matplotlib.pyplot.close: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.close.html

선수별 속도 벡터 및 속력 산출¶

In [1]:

import os
new_dir = '/Users/limjongjun/Desktop/JayJay/Growth/Python/soccer-analytics'
os.chdir(new_dir)

import warnings
import numpy as np
import pandas as pd
import scipy.signal as signal
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from src.plot_utils import draw_pitch

warnings.simplefilter(action="ignore", category=pd.errors.PerformanceWarning)

(1) 결합 데이터 불러오기¶

In [2]:

match_id = 1
file = f'data_metrica/data/Sample_Game_{match_id}/Sample_Game_{match_id}_IntegratedData.csv'
traces = pd.read_csv(file, header=0, index_col=0)
traces

Out[2]:

	period	time	H11_x	H11_y	H01_x	H01_y	H02_x	H02_y	H03_x	H03_y	...	A24_speed	A26_vx	A26_vy	A26_speed	A27_vx	A27_vy	A27_speed	A28_vx	A28_vy	A28_speed
frame
1	1	0.04	0.08528	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	1	0.08	0.09984	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	1	0.12	0.11856	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	1	0.16	0.12584	32.80184	33.92688	44.41556	35.03448	33.31184	32.18176	24.17672	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
5	1	0.20	0.13416	32.80184	33.90088	44.38292	35.01056	33.33224	32.18592	24.15904	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
145002	2	5800.08	9.76144	37.15860	NaN	NaN	68.47672	24.07336	71.52288	22.09116	...	0.499132	0.354571	0.405385	0.538570	0.298571	-0.494308	0.577482	0.0	0.0	0.0
145003	2	5800.12	9.82800	37.15860	NaN	NaN	68.50792	24.08356	71.52080	22.08028	...	0.499132	0.453857	0.299462	0.543750	0.359857	-0.606769	0.705455	0.0	0.0	0.0
145004	2	5800.16	9.88832	37.15860	NaN	NaN	68.54744	24.09308	71.53744	22.09932	...	0.499132	0.553143	0.193538	0.586024	0.421143	-0.719231	0.833459	0.0	0.0	0.0
145005	2	5800.20	9.92576	37.15792	NaN	NaN	68.58176	24.10464	71.58216	22.12176	...	0.499132	0.652429	0.087615	0.658285	0.482429	-0.831692	0.961483	0.0	0.0	0.0
145006	2	5800.24	9.92576	37.15792	NaN	NaN	68.64416	24.11008	71.63312	22.07824	...	0.499132	0.751714	-0.018308	0.751937	0.543714	-0.944154	1.089519	0.0	0.0	0.0

145006 rows × 147 columns

(2) 특정 선수 속도/속력 산출¶

In [3]:

period = 1
player = 'H11' # 11번선수의 속도 벡터만 추출
player_cols = ['period', 'time', f'{player}_x', f'{player}_y']
player_traces = traces.loc[traces['period'] == period, player_cols]
player_traces

Out[3]:

	period	time	H11_x	H11_y
frame
1	1	0.04	0.08528	32.80184
2	1	0.08	0.09984	32.80184
3	1	0.12	0.11856	32.80184
4	1	0.16	0.12584	32.80184
5	1	0.20	0.13416	32.80184
...	...	...	...	...
71264	1	2850.56	15.91304	45.93332
71265	1	2850.60	15.93800	45.93332
71266	1	2850.64	15.95360	45.93332
71267	1	2850.68	15.96920	45.93332
71268	1	2850.72	15.96920	45.93332

71268 rows × 4 columns

In [4]:

dt = player_traces['time'].diff()
vx = player_traces[f'{player}_x'].diff() / dt
vy = player_traces[f'{player}_y'].diff() / dt

raw_speeds = np.sqrt(vx ** 2 + vy ** 2) # 속력 구하기

# 12 m/s 보다 큰 값은 사람이 낼 수 없는 속력으로 이상치 처리
lim_speed = 12
vx[raw_speeds > lim_speed] = np.nan
vy[raw_speeds > lim_speed] = np.nan

plt.figure(figsize=(15, 5))
plt.rcParams.update({'font.size': 15})
plt.plot(player_traces['time'][10000:12000], raw_speeds[10000:12000])
plt.xlabel('Time [s]')
plt.ylabel('Speed [m/s]')
plt.show()

(3) 사비츠키-골레이 필터(Savitzky-Golay filter)를 활용한 속도/속력 스무딩(smoothing)¶

In [5]:

vx = signal.savgol_filter(vx, window_length=13, polyorder=1) # 주어진 점 좌우 13개 데이터를 참고
vy = signal.savgol_filter(vy, window_length=13, polyorder=1)
speeds = np.sqrt(vx ** 2 + vy ** 2)

plt.figure(figsize=(15, 5))
plt.rcParams.update({'font.size': 15})
plt.plot(player_traces['time'][10000:12000], speeds[10000:12000])
plt.xlabel('Time [s]')
plt.ylabel('Speed [m/s]')
plt.show()

(4) 선수별 속도/속력 산출 함수 구현¶

In [6]:

def calc_running_features(traces, lim_speed=12, smoothing=True, window_length=13, polyorder=1):
    players = [c[:-2] for c in traces.columns if c[0] in ['H', 'A'] and c.endswith('_x')]

    for period in traces['period'].unique():
        period_traces = traces[traces['period'] == period]

        idx = period_traces.index
        dt = period_traces['time'].diff()

        for player in players:
            vx = period_traces[f'{player}_x'].diff() / dt
            vy = period_traces[f'{player}_y'].diff() / dt
            
            raw_speeds = np.sqrt(vx ** 2 + vy ** 2)
            vx[raw_speeds > lim_speed] = np.nan
            vy[raw_speeds > lim_speed] = np.nan
            vx = vx.interpolate()
            vy = vy.interpolate()

            if smoothing:
                vx = signal.savgol_filter(vx, window_length=13, polyorder=1)
                vy = signal.savgol_filter(vy, window_length=13, polyorder=1)
            
            traces.loc[idx, f'{player}_vx'] = vx
            traces.loc[idx, f'{player}_vy'] = vy
            traces.loc[idx, f'{player}_speed'] = np.sqrt(vx ** 2 + vy ** 2)

    return traces

In [7]:

traces = calc_running_features(traces)
traces

Out[7]:

	period	time	H11_x	H11_y	H01_x	H01_y	H02_x	H02_y	H03_x	H03_y	...	A24_speed	A26_vx	A26_vy	A26_speed	A27_vx	A27_vy	A27_speed	A28_vx	A28_vy	A28_speed
frame
1	1	0.04	0.08528	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
2	1	0.08	0.09984	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
3	1	0.12	0.11856	32.80184	33.95392	44.41896	35.04904	33.22684	32.16408	24.15972	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
4	1	0.16	0.12584	32.80184	33.92688	44.41556	35.03448	33.31184	32.18176	24.17672	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
5	1	0.20	0.13416	32.80184	33.90088	44.38292	35.01056	33.33224	32.18592	24.15904	...	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
145002	2	5800.08	9.76144	37.15860	NaN	NaN	68.47672	24.07336	71.52288	22.09116	...	0.499132	0.354571	0.405385	0.538570	0.298571	-0.494308	0.577482	0.0	0.0	0.0
145003	2	5800.12	9.82800	37.15860	NaN	NaN	68.50792	24.08356	71.52080	22.08028	...	0.499132	0.453857	0.299462	0.543750	0.359857	-0.606769	0.705455	0.0	0.0	0.0
145004	2	5800.16	9.88832	37.15860	NaN	NaN	68.54744	24.09308	71.53744	22.09932	...	0.499132	0.553143	0.193538	0.586024	0.421143	-0.719231	0.833459	0.0	0.0	0.0
145005	2	5800.20	9.92576	37.15792	NaN	NaN	68.58176	24.10464	71.58216	22.12176	...	0.499132	0.652429	0.087615	0.658285	0.482429	-0.831692	0.961483	0.0	0.0	0.0
145006	2	5800.24	9.92576	37.15792	NaN	NaN	68.64416	24.11008	71.63312	22.07824	...	0.499132	0.751714	-0.018308	0.751937	0.543714	-0.944154	1.089519	0.0	0.0	0.0

145006 rows × 147 columns

In [8]:

traces.to_csv('data_metrica/data/Sample_Game_1/Sample_Game_1_IntegratedData.csv')

(5) 여러 경기 위치 추적 데이터 가공 및 저장¶

In [9]:

matches = [d for d in os.listdir('data_metrica') if not d.startswith('.')]
matches.sort()
matches

Out[9]:

['README.md', 'data', 'documentation']

In [10]:

for match in matches:
    home_file = f'data_metrica/data/{match}/{match}_RawTrackingData_Home_Team.csv'
    away_file = f'data_metrica/data/{match}/{match}_RawTrackingData_Away_Team.csv'
    event_file = f'data_metrica/data/{match}/{match}_RawEventsData.csv'
    try:
        home_traces = pd.read_csv(home_file, header=[0, 1, 2])
        away_traces = pd.read_csv(away_file, header=[0, 1, 2])
        events = pd.read_csv(event_file, header=0)
    except FileNotFoundError:
        continue

    home_players = [c[2] for c in home_traces.columns[3:-2:2]]
    home_trace_cols = [[f'H{int(p[6:]):02d}_x', f'H{int(p[6:]):02d}_y'] for p in home_players]
    home_trace_cols = np.array(home_trace_cols).flatten().tolist()
    home_traces.columns = ['period', 'frame', 'time'] + home_trace_cols + ['ball_x', 'ball_y']
    home_traces = home_traces.set_index('frame').astype(float)
    home_traces['period'] = home_traces['period'].astype(int)

    away_players = [c[2] for c in away_traces.columns[3:-2:2]]
    away_trace_cols = [[f'A{int(p[6:]):02d}_x', f'A{int(p[6:]):02d}_y'] for p in away_players]
    away_trace_cols = np.array(away_trace_cols).flatten().tolist()
    away_traces.columns = ['period', 'frame', 'time'] + away_trace_cols + ['ball_x', 'ball_y']
    away_traces = away_traces.set_index('frame').astype(float)
    away_traces['period'] = away_traces['period'].astype(int)
    
    cols = home_traces.columns[:-2].tolist() + away_traces.columns[2:].tolist()
    traces = pd.merge(home_traces, away_traces)[cols]
    traces.index = home_traces.index.astype(int)

    x_cols = [c for c in traces.columns if c.endswith('_x')]
    y_cols = [c for c in traces.columns if c.endswith('_y')]
    traces.loc[traces['period'] == 2, x_cols + y_cols] = 1 - traces.loc[traces['period'] == 2, x_cols + y_cols]
    traces[x_cols] *= 104
    traces[y_cols] *= 68

    events.loc[events['Subtype'].isna(), 'Subtype'] = events.loc[events['Subtype'].isna(), 'Type']
    for i, event in events.iterrows():
        start_frame = event['Start Frame']
        end_frame = event['End Frame']
        traces.loc[start_frame:end_frame-1, 'event_player'] = event['From']
        traces.loc[start_frame:end_frame-1, 'event_type'] = event['Type']
        traces.loc[start_frame:end_frame-1, 'event_subtype'] = event['Subtype']
    
    traces = calc_running_features(traces)
    traces.to_csv(f'data_metrica/{match}/{match}_IntegratedData.csv')
    print(f'Integrated data saved for {match}.')

경기 장면 애니메이션 시각화¶

(1) 속도 벡터 포함 특정 시점 이미지 시각화¶

In [11]:

frame = 1000

file = f'data_metrica/data/Sample_Game_{match_id}/Sample_Game_{match_id}_IntegratedData.csv'
traces = pd.read_csv(file, header=0, index_col=0)
data = traces.loc[frame]

fig, ax = draw_pitch(pitch='white', line='black')

for team, color in zip(['H', 'A'], ['r', 'b']):
    x_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_x')]
    y_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_y')]
    ax.scatter(data[x_cols], data[y_cols], s=100, c=color, alpha=0.7)
    
    for x, y in zip(x_cols, y_cols):
        if not (np.isnan(data[x]) or np.isnan(data[y])):
            ax.text(data[x] + 0.5, data[y] + 0.5, int(x[1:3]), fontsize=13, color=color)
    
    vx_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_vx')]
    vy_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_vy')]
    # quiver : 화살표를 한번에 시각화 
    ax.quiver(
        data[x_cols].astype(float), data[y_cols].astype(float),
        data[vx_cols].astype(float), data[vy_cols].astype(float),
        color=color, scale=8, scale_units='inches', width=0.002, alpha=0.7
    )

ax.scatter(data['ball_x'], data['ball_y'], s=80, color='w', edgecolors='k')

time_text = f"{int(data['time'] // 60):02d}:{data['time'] % 60:05.2f}"
if not pd.isnull(data['event_subtype']):
    event_text = f"{data['event_subtype']} by {data['event_player']}" 
else:
    event_text = ''
ax.text(51, 67, time_text, fontsize=15, ha='right', va='top')
ax.text(53, 67, event_text, fontsize=15, ha='left', va='top')
    
plt.show()

(2) 시점별 이미지 시각화 함수 구현¶

In [12]:

def plot_snapshot(
    data, figax=None, team_colors=('r', 'b'),
    annotate_players=True, annotate_events=True, show_velocities=True
):
    if figax is None:
        fig, ax = draw_pitch(pitch='white', line='black')
    else:
        fig, ax = figax

    figobjs = []
    for team, color in zip(['H', 'A'], team_colors):
        x_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_x')]
        y_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_y')]
        obj = ax.scatter(data[x_cols], data[y_cols], s=100, c=color, alpha=0.7)
        figobjs.append(obj)

        if show_velocities:
            vx_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_vx')]
            vy_cols = [c for c in data.keys() if c.startswith(team) and c.endswith('_vy')]
            obj = ax.quiver(
                data[x_cols].astype(float), data[y_cols].astype(float),
                data[vx_cols].astype(float), data[vy_cols].astype(float),
                color=color, scale=8, scale_units='inches', width=0.002, alpha=0.7
            )
            figobjs.append(obj)
        
        if annotate_players:
            for x, y in zip(x_cols, y_cols):
                if not (np.isnan(data[x]) or np.isnan(data[y])):
                    obj = ax.text(data[x] + 0.5, data[y] + 0.5, int(x[1:3]), fontsize=13, color=color)
                    figobjs.append(obj)

    time_text = f"{int(data['time'] // 60):02d}:{data['time'] % 60:05.2f}"
    if annotate_events:
        if not pd.isnull(data['event_subtype']):
            event_text = f"{data['event_subtype']} by {data['event_player']}" 
        else:
            event_text = ''
        figobjs.append(ax.text(51, 67, time_text, fontsize=15, ha='right', va='top'))
        figobjs.append(ax.text(53, 67, event_text, fontsize=15, ha='left', va='top'))
    else:
        figobjs.append(ax.text(52, 67, time_text, fontsize=15, ha='center', va='top'))
        
    obj = ax.scatter(data['ball_x'], data['ball_y'], s=80, color='w', edgecolors='k')
    figobjs.append(obj)

    ax.set_xlim(-10, 114)
    ax.set_ylim(-7, 75)
    
    return fig, ax, figobjs

In [13]:

frame = 1000
fig, ax, figobjs = plot_snapshot(traces.loc[frame])
plt.show()

figobjs

Out[13]:

[<matplotlib.collections.PathCollection at 0x16924a040>,
 <matplotlib.quiver.Quiver at 0x168687c70>,
 Text(5.34744, 34.50204, '11'),
 Text(21.41024, 56.25456, '1'),
 Text(20.68432, 47.09428, '2'),
 Text(20.99112, 39.804680000000005, '3'),
 Text(20.77688, 28.48608, '4'),
 Text(37.81416, 64.02968, '5'),
 Text(32.980239999999995, 52.19632, '6'),
 Text(28.1172, 41.54004, '7'),
 Text(34.924, 25.87148, '8'),
 Text(55.49832, 30.0392, '9'),
 Text(44.2112, 43.02924, '10'),
 <matplotlib.collections.PathCollection at 0x169231640>,
 <matplotlib.quiver.Quiver at 0x1691eae80>,
 Text(83.18520000000001, 34.36468, '25'),
 Text(55.11976000000001, 28.17328, '15'),
 Text(56.21592, 40.1678, '16'),
 Text(50.58536, 58.67876, '17'),
 Text(29.66368, 18.139200000000002, '18'),
 Text(34.2376, 37.83744, '19'),
 Text(42.56384, 54.5906, '20'),
 Text(17.01312, 57.3038, '21'),
 Text(37.09448, 66.29136, '22'),
 Text(22.0072, 42.10988, '23'),
 Text(24.42312, 45.04952, '24'),
 Text(51, 67, '00:40.00'),
 Text(53, 67, 'PASS by Player22'),
 <matplotlib.collections.PathCollection at 0x169251490>]

(3) 경기 장면 애니메이션 시각화 함수 구현¶

In [24]:

def save_clip(
    clip_traces, fname='test', fps=25, figax=None, team_colors=('r', 'b'),
    annotate_players=True, annotate_events=True, show_velocities=True
):
    metadata = dict(title='Tracking Data', artist='Matplotlib', comment='Metrica tracking data clip')
    
    # FFMpegWriter 패키지 필요
    writer = animation.FFMpegWriter(fps=fps, metadata=metadata) 

    if not os.path.exists('match_clips'):
        os.makedirs('match_clips')
    file = f'/Users/limjongjun/Desktop/JayJay/Growth/Python/soccer-analytics/data_metrica/match_clips/{fname}.mp4'

    if figax is None:
        fig, ax = draw_pitch(pitch='white', line='black')
    else:
        fig, ax = figax
    fig.set_tight_layout(True)

    with writer.saving(fig, file, dpi=100):
        for i in clip_traces.index:
            frame_data = clip_traces.loc[i]
            fig, ax, figobjs = plot_snapshot(
                frame_data, (fig, ax), team_colors,
                annotate_players, annotate_events, show_velocities
            )
            writer.grab_frame()

            # 한 장면 장면 시각화 할 때마다 메모리에 객체들이 쌓이지 않도록 remove
            for obj in figobjs:
                obj.remove()

    plt.clf()
    plt.close(fig)

In [25]:

clip_traces = traces[:1000]
save_clip(clip_traces)

---------------------------------------------------------------------------
FileNotFoundError                         Traceback (most recent call last)
Cell In[25], line 2
      1 clip_traces = traces[:1000]
----> 2 save_clip(clip_traces)

Cell In[24], line 20, in save_clip(clip_traces, fname, fps, figax, team_colors, annotate_players, annotate_events, show_velocities)
     17     fig, ax = figax
     18 fig.set_tight_layout(True)
---> 20 with writer.saving(fig, file, dpi=100):
     21     for i in clip_traces.index:
     22         frame_data = clip_traces.loc[i]

File ~/opt/anaconda3/envs/class101/lib/python3.8/contextlib.py:113, in _GeneratorContextManager.__enter__(self)
    111 del self.args, self.kwds, self.func
    112 try:
--> 113     return next(self.gen)
    114 except StopIteration:
    115     raise RuntimeError("generator didn't yield") from None

File ~/opt/anaconda3/envs/class101/lib/python3.8/site-packages/matplotlib/animation.py:231, in AbstractMovieWriter.saving(self, fig, outfile, dpi, *args, **kwargs)
    225 """
    226 Context manager to facilitate writing the movie file.
    227 
    228 ``*args, **kw`` are any parameters that should be passed to `setup`.
    229 """
    230 # This particular sequence is what contextlib.contextmanager wants
--> 231 self.setup(fig, outfile, dpi, *args, **kwargs)
    232 try:
    233     yield self

File ~/opt/anaconda3/envs/class101/lib/python3.8/site-packages/matplotlib/animation.py:320, in MovieWriter.setup(self, fig, outfile, dpi)
    317 self._w, self._h = self._adjust_frame_size()
    318 # Run here so that grab_frame() can write the data to a pipe. This
    319 # eliminates the need for temp files.
--> 320 self._run()

File ~/opt/anaconda3/envs/class101/lib/python3.8/site-packages/matplotlib/animation.py:330, in MovieWriter._run(self)
    327 _log.info('MovieWriter._run: running command: %s',
    328           cbook._pformat_subprocess(command))
    329 PIPE = subprocess.PIPE
--> 330 self._proc = subprocess.Popen(
    331     command, stdin=PIPE, stdout=PIPE, stderr=PIPE,
    332     creationflags=subprocess_creation_flags)

File ~/opt/anaconda3/envs/class101/lib/python3.8/subprocess.py:858, in Popen.__init__(self, args, bufsize, executable, stdin, stdout, stderr, preexec_fn, close_fds, shell, cwd, env, universal_newlines, startupinfo, creationflags, restore_signals, start_new_session, pass_fds, encoding, errors, text)
    854         if self.text_mode:
    855             self.stderr = io.TextIOWrapper(self.stderr,
    856                     encoding=encoding, errors=errors)
--> 858     self._execute_child(args, executable, preexec_fn, close_fds,
    859                         pass_fds, cwd, env,
    860                         startupinfo, creationflags, shell,
    861                         p2cread, p2cwrite,
    862                         c2pread, c2pwrite,
    863                         errread, errwrite,
    864                         restore_signals, start_new_session)
    865 except:
    866     # Cleanup if the child failed starting.
    867     for f in filter(None, (self.stdin, self.stdout, self.stderr)):

File ~/opt/anaconda3/envs/class101/lib/python3.8/subprocess.py:1704, in Popen._execute_child(self, args, executable, preexec_fn, close_fds, pass_fds, cwd, env, startupinfo, creationflags, shell, p2cread, p2cwrite, c2pread, c2pwrite, errread, errwrite, restore_signals, start_new_session)
   1702     if errno_num != 0:
   1703         err_msg = os.strerror(errno_num)
-> 1704     raise child_exception_type(errno_num, err_msg, err_filename)
   1705 raise child_exception_type(err_msg)

FileNotFoundError: [Errno 2] No such file or directory: 'ffmpeg'