Adding a new config for training on UBFC-Phys

[zzl0318]

I tried a config like this, but the results was very pore. Is their something wrong with my config file?
Calculating metrics!
100%|███████████████████████████████████████████| 42/42 [00:30<00:00, 1.38it/s]
FFT MAE (FFT Label): 34.005301339285715 +/- 3.033595794199921
FFT RMSE (FFT Label): 39.27943563449327 +/- 14.6103528726731
FFT MAPE (FFT Label): 31.529103551082756 +/- 2.462837199273431
FFT Pearson (FFT Label): -0.07576602068851186 +/- 0.15765940426351263
FFT SNR (FFT Label): -12.177335893780446 +/- 1.1786350898544593 (dB)

BASE: ['']
TOOLBOX_MODE: "train_and_test" # "train_and_test" or "only_test"
TRAIN:
BATCH_SIZE: 4
EPOCHS: 30
LR: 9e-3
MODEL_FILE_NAME: PURE_PURE_UBFC_tscan
PLOT_LOSSES_AND_LR: True
DATA:
FILTERING:
USE_EXCLUSION_LIST: True
EXCLUSION_LIST: [
's3_T1', 's8_T1', 's9_T1', 's26_T1', 's28_T1', 's30_T1', 's31_T1', 's32_T1',
's33_T1', 's40_T1', 's52_T1', 's53_T1', 's54_T1', 's56_T1', 's1_T2', 's4_T2',
's6_T2', 's8_T2', 's9_T2', 's11_T2', 's12_T2', 's13_T2', 's14_T2', 's19_T2',
's21_T2', 's22_T2', 's25_T2', 's26_T2', 's27_T2', 's28_T2', 's31_T2', 's32_T2',
's33_T2', 's35_T2', 's38_T2', 's39_T2', 's41_T2', 's42_T2', 's45_T2', 's47_T2',
's48_T2', 's52_T2', 's53_T2', 's55_T2', 's5_T3', 's8_T3', 's9_T3', 's10_T3',
's13_T3', 's14_T3', 's17_T3', 's22_T3', 's25_T3', 's26_T3', 's28_T3', 's30_T3',
's32_T3', 's33_T3', 's35_T3', 's37_T3', 's40_T3', 's47_T3', 's48_T3', 's49_T3',
's50_T3', 's52_T3', 's53_T3']
SELECT_TASKS: True
TASK_LIST: ['T1', 'T2', 'T3']
FS: 35
DATASET: UBFC-PHYS
DO_PREPROCESS: True # if first time, should be true
DATA_FORMAT: NDCHW
DATA_PATH: "/mnt/newdisk/rppg_datasets/UBFC-Phys-raw" # Raw dataset path, need to be updated
CACHED_PATH: "/home/zzl/rPPG-Toolbox/dataset/UBFC-Pyhs-preprocessed" # Processed dataset save path, need to be updated
EXP_DATA_NAME: ""
BEGIN: 0.0
END: 0.8
PREPROCESS:
DATA_TYPE: ['DiffNormalized','Standardized']
DATA_AUG: ['None'] # 'None' or 'Motion' is supported, used if the data path points to an augmented dataset or requires augmentation
LABEL_TYPE: DiffNormalized
DO_CHUNK: True
CHUNK_LENGTH: 180
CROP_FACE:
DO_CROP_FACE: True
BACKEND: 'HC' # HC for Haar Cascade, RF for RetinaFace
USE_LARGE_FACE_BOX: True
LARGE_BOX_COEF: 1.5
DETECTION:
DO_DYNAMIC_DETECTION: False
DYNAMIC_DETECTION_FREQUENCY: 30
USE_MEDIAN_FACE_BOX: False # This should be used ONLY if dynamic detection is used
RESIZE:
H: 72
W: 72
VALID:
DATA:
FS: 30
DATASET: UBFC-PHYS
DO_PREPROCESS: False # if first time, should be true
DATA_FORMAT: NDCHW
DATA_PATH: "/mnt/newdisk/rppg_datasets/UBFC-Phys-raw" # Raw dataset path, need to be updated
CACHED_PATH: "/home/zzl/rPPG-Toolbox/dataset/UBFC-Pyhs-preprocessed" # Processed dataset save path, need to be updated
EXP_DATA_NAME: ""
BEGIN: 0.8
END: 1.0
PREPROCESS:
DATA_TYPE: [ 'DiffNormalized','Standardized' ]
DATA_AUG: ['None'] # 'None' or 'Motion' is supported, used if the data path points to an augmented dataset or requires augmentation
LABEL_TYPE: DiffNormalized
DO_CHUNK: True
CHUNK_LENGTH: 180
CROP_FACE:
DO_CROP_FACE: True
BACKEND: 'HC' # HC for Haar Cascade, RF for RetinaFace
USE_LARGE_FACE_BOX: True
LARGE_BOX_COEF: 1.5
DETECTION:
DO_DYNAMIC_DETECTION: False
DYNAMIC_DETECTION_FREQUENCY : 30
USE_MEDIAN_FACE_BOX: False # This should be used ONLY if dynamic detection is used
RESIZE:
H: 72
W: 72
TEST:
METRICS: ['MAE', 'RMSE', 'MAPE', 'Pearson', 'SNR', 'BA']
USE_LAST_EPOCH: True # to use provided validation dataset to find the best epoch, should be false
DATA:
FS: 30
DATASET: UBFC-rPPG
DO_PREPROCESS: True # if first time, should be true
DATA_FORMAT: NDCHW
DATA_PATH: "/mnt/newdisk/rppg_datasets/UBFC-rPPG-raw" # need to be updated
CACHED_PATH: "/home/zzl/rPPG-Toolbox/dataset/ubfc-rppg-preprocessed" # Processed dataset save path, need to be updated
EXP_DATA_NAME: ""
BEGIN: 0.0
END: 1.0
PREPROCESS:
DATA_TYPE: [ 'DiffNormalized','Standardized' ]
LABEL_TYPE: DiffNormalized
DO_CHUNK: True
CHUNK_LENGTH: 180
CROP_FACE:
DO_CROP_FACE: True
BACKEND: 'HC' # HC for Haar Cascade, RF for RetinaFace
USE_LARGE_FACE_BOX: True
LARGE_BOX_COEF: 1.5
DETECTION:
DO_DYNAMIC_DETECTION: False
DYNAMIC_DETECTION_FREQUENCY: 30
USE_MEDIAN_FACE_BOX: False # This should be used ONLY if dynamic detection is used
RESIZE:
H: 72
W: 72
DEVICE: cuda:0
NUM_OF_GPU_TRAIN: 1
LOG:
PATH: runs/exp
MODEL:
DROP_RATE: 0.2
NAME: Tscan
TSCAN:
FRAME_DEPTH: 10
INFERENCE:
BATCH_SIZE: 4
EVALUATION_METHOD: "FFT" # "FFT" or "peak detection"
EVALUATION_WINDOW:
USE_SMALLER_WINDOW: False # Change this if you'd like an evaluation window smaller than the test video length
WINDOW_SIZE: 10 # In seconds
MODEL_PATH: ""

[yahskapar]

Hi @zzl0318,

It's totally possible training on UBFC-Phys is quite unstable, personally I've never even tried and only used it as an evaluation dataset in the past. Try tweaking things like using RetinaFace instead of Haar Cascade for the face detection backend, as well as enabling dynamic detection and utilizing a median face box. UBFC-Phys has quite a bit of rigid head motion that can be significantly influenced by face detection-related parameters.

If that doesn't work at all, it's also worth playing around with the more general hyperparameters (e.g., learning rate, epochs) as well.

All the best,

Akshay

[yahskapar]

I'll go ahead and close this due to the lack of further discussion, but feel free to re-open or create a new one if you have any other questions or concerns @zzl0318. Thanks for using this toolbox!