import pandas as pd import json from os.path import exists def read_derivedAttr(ar): derivedFilename = ar+"/derived_variables/derived_output.csv" if not exists(derivedFilename): return pd.DataFrame() derived_df = pd.read_csv(derivedFilename) facial_cols = [col for col in derived_df if "fac_" in col] acoustic_cols = [col for col in derived_df if "aco_" in col] movement_cols = [col for col in derived_df if "mov_" in col] nlp_cols = [col for col in derived_df if "nlp_" in col] derived_facial = derived_df.loc[:,derived_df.columns.isin(facial_cols)] derived_acoustic = derived_df.loc[:,derived_df.columns.isin(acoustic_cols)] derived_movement = derived_df.loc[:,derived_df.columns.isin(movement_cols)] derived_nlp = derived_df.loc[:,derived_df.columns.isin(nlp_cols)] ids = derived_df["Filename"].tolist() return {"ids": ids, "facialAttr": facial_cols, "acousticAttr": acoustic_cols, "movementAttr": movement_cols, "speechAttr": nlp_cols} def read_medatada(ar, ar2): metaData={} metdataFilename = ar+"/"+ar2 if exists(metdataFilename): metadataDf = pd.read_csv(metdataFilename) metaData = metadataDf.to_json(orient="records") return metaData def read_derivedDf(ar): derivedFilename = ar+"/derived_variables/derived_output.csv" if not exists(derivedFilename): return pd.DataFrame() derived_df = pd.read_csv(derivedFilename) return derived_df def read_rawFacialDf(ar, id): skip_cols = ["frame", "face_id", "error_reason", "timestamp", "confidence", "success", "dbm_master_url", "error_reason", " confidence", " face_id", " success", " timestamp", "s_confidence", ] facial_asym_filename = ar + "/raw_variables/"+id+"/facial/face_asymmetry/"+id+"_facasym.csv" facial_au_filename = ar + "/raw_variables/"+id+"/facial/face_au/"+id+"_facau.csv" facial_expr_filename = ar + "/raw_variables/"+id+"/facial/face_expressivity/"+id+"_facemo.csv" landmark_filename = ar + "/raw_variables/"+id+"/facial/face_landmark/"+id+"_faclmk.csv" if not exists(facial_asym_filename) or not exists(facial_au_filename) or not exists(facial_expr_filename): return pd.DataFrame() facial_asym = pd.read_csv(facial_asym_filename) facial_asym_cols = [col for col in facial_asym if col not in skip_cols] facial_au = pd.read_csv(facial_au_filename) facial_au_cols = [col for col in facial_au if col not in skip_cols] facial_expr = pd.read_csv(facial_expr_filename) facial_expr_cols = [col for col in facial_expr if col not in skip_cols and "AU" not in col] face_df = facial_asym.loc[:, ~facial_asym.columns.isin(skip_cols)].copy() for el in facial_au_cols: face_df[el] = facial_au[el] for el in facial_expr_cols: face_df[el] = facial_expr[el] return face_df.fillna(0) def read_rawMovementDf(ar, id): skip_cols = ["error_reason", "dbm_master_url", "Frames", " Frames"] gaze_filename = ar + "/raw_variables/"+id+"/movement/gaze/"+id+"_eyegaze.csv" head_movement_filename = ar + "/raw_variables/"+id+"/movement/head_movement/"+id+"_headmov.csv" head_pose_filename = ar + "/raw_variables/"+id+"/movement/head_pose/"+id+"_headpose.csv" blinks_filename = ar + "/raw_variables/"+id+"/movement/eye_blink/"+id+"_eyeblinks.csv" fac_tremor_filename = ar + "/raw_variables/"+id+"/movement/facial_tremor/"+id+"_fac_tremor.csv" voice_tremor_filename = ar + "/raw_variables/"+id+"/movement/voice_tremor/"+id+"_vtremor.csv" if not exists(gaze_filename) or not exists(head_movement_filename) or not exists(head_pose_filename): return pd.DataFrame() gaze = pd.read_csv(gaze_filename) gaze_cols = [col for col in gaze if col not in skip_cols] head_movement = pd.read_csv(head_movement_filename) head_movement_cols = [col for col in head_movement if col not in skip_cols] # blinks = pd.read_csv(blinks_filename) # blinks_cols = [col for col in blinks if col not in skip_cols] # fac_tremor = pd.read_csv(fac_tremor_filename) # fac_tremor_cols = [col for col in fac_tremor if col not in skip_cols] voice_tremor = pd.read_csv(voice_tremor_filename) voice_tremor_cols = [col for col in voice_tremor if col not in skip_cols] head_pose = pd.read_csv(head_pose_filename) head_pose_cols = [col for col in head_pose if col not in skip_cols] movement_df = head_pose.loc[:, ~head_pose.columns.isin(skip_cols)].copy() for el in head_movement_cols: movement_df[el] = head_movement[el] for el in gaze_cols: movement_df[el] = gaze[el] # for el in blinks_cols: # movement_df[el] = blinks[el] # for el in fac_tremor_cols: # movement_df[el] = fac_tremor[el] for el in voice_tremor_cols: movement_df[el] = voice_tremor[el] return movement_df.fillna(0) def read_rawAcousticDf(ar, id): skip_cols = ["error_reason", "dbm_master_url", "Frames", " Frames", "aco_voicelabel"] fm_filename = ar + "/raw_variables/"+id+"/acoustic/formant_freq/"+id+"_formant.csv" gne_filename = ar + "/raw_variables/"+id+"/acoustic/glottal_noise/"+id+"_gne.csv" hnr_filename = ar + "/raw_variables/"+id+"/acoustic/harmonic_noise/"+id+"_hnr.csv" intt_filename = ar + "/raw_variables/"+id+"/acoustic/intensity/"+id+"_intensity.csv" mfcc_filename = ar + "/raw_variables/"+id+"/acoustic/mfcc/"+id+"_mfcc.csv" pitch_filename = ar + "/raw_variables/"+id+"/acoustic/pitch/"+id+"_pitch.csv" jitter_filename = ar + "/raw_variables/"+id+"/acoustic/jitter/"+id+"_jitter.csv" pause_segment_filename = ar + "/raw_variables/"+id+"/acoustic/pause_segment/"+id+"_pausechar.csv" shimmer_filename = ar + "/raw_variables/"+id+"/acoustic/shimmer/"+id+"_shimmer.csv" voice_frame_score_filename = ar + "/raw_variables/"+id+"/acoustic/voice_frame_score/"+id+"_voiceprev.csv" filename_list = [fm_filename, gne_filename, hnr_filename, intt_filename,mfcc_filename, pitch_filename, pause_segment_filename, shimmer_filename, voice_frame_score_filename] if not exists(fm_filename) or not exists(gne_filename) or not exists(hnr_filename) or not exists(intt_filename) or not exists(mfcc_filename) or not exists(pitch_filename): return pd.DataFrame fm = pd.read_csv(fm_filename) fm_cols = [col for col in fm if col not in skip_cols] gne = pd.read_csv(gne_filename) gne_cols = [col for col in gne if col not in skip_cols] hnr = pd.read_csv(hnr_filename) hnr_cols = [col for col in hnr if col not in skip_cols] intt = pd.read_csv(intt_filename) intt_cols = [col for col in intt if col not in skip_cols] mfcc = pd.read_csv(mfcc_filename) mfcc_cols = [col for col in mfcc if col not in skip_cols] pitch = pd.read_csv(pitch_filename) pitch_cols = [col for col in pitch if col not in skip_cols] pause_segment = pd.read_csv(pause_segment_filename) pause_segment_cols = [col for col in pause_segment if col not in skip_cols] jitter = pd.read_csv(jitter_filename) jitter_cols = [col for col in jitter if col not in skip_cols] shimmer = pd.read_csv(shimmer_filename) shimmer_cols = [col for col in shimmer if col not in skip_cols] voice_frame_score = pd.read_csv(voice_frame_score_filename) voice_frame_score_cols = [col for col in voice_frame_score if col not in skip_cols] acoustic_df = fm.loc[:, ~fm.columns.isin(skip_cols)].copy() for el in gne_cols: acoustic_df[el] = gne[el] for el in hnr_cols: acoustic_df[el] = hnr[el] for el in intt_cols: acoustic_df[el] = intt[el] for el in mfcc_cols: acoustic_df[el] = mfcc[el] for el in pitch_cols: acoustic_df[el] = pitch[el] for el in pause_segment_cols: acoustic_df[el] = pause_segment[el] for el in jitter_cols: acoustic_df[el] = jitter[el] for el in shimmer_cols: acoustic_df[el] = shimmer[el] for el in voice_frame_score_cols: acoustic_df[el] = voice_frame_score[el] return acoustic_df.fillna(0) def load(): global videoIds read_derivedAttr() read_derivedDf() if __name__=="__main__": load()