Hi there!
I need an advice.
I have a task to count the passangeers in the bus. I have a camera looking vertically down at the door place. (Pictures attached)
I have my traned model with a single class. So, in my opinion, I can neglect such metrics as BoxLost, Class Lost and DFL Lost (Am I right?)
My training arguments is
task: detect
mode: train
model: /content/drive/MyDrive/YOLO_detection/trainCfg-1000-16_DayPlusNight_mozaic_Y11s_runs/train/weights/last.pt
data: /content/data.yaml
epochs: 1000
time: null
patience: 50
batch: 30
imgsz: 640
save: true
save_period: 50
cache: disk
device: null
workers: 32
project: /content/drive/MyDrive/YOLO_detection/trainCfg-1000-16_DayPlusNight_mozaic_Y11s_runs
name: train
exist_ok: false
pretrained: yolo11s.pt
optimizer: auto
verbose: true
seed: 0
deterministic: true
single_cls: true
rect: false
cos_lr: true
close_mosaic: 10
resume: /content/drive/MyDrive/YOLO_detection/trainCfg-1000-16_DayPlusNight_mozaic_Y11s_runs/train/weights/last.pt
amp: true
fraction: 1.0
profile: false
freeze: null
multi_scale: false
compile: false
overlap_mask: true
mask_ratio: 4
dropout: 0.0
val: true
split: val
save_json: false
conf: null
iou: 0.7
max_det: 300
half: false
dnn: false
plots: true
source: null
vid_stride: 1
stream_buffer: false
visualize: false
augment: true
agnostic_nms: false
classes: null
retina_masks: false
embed: null
show: false
save_frames: false
save_txt: false
save_conf: false
save_crop: false
show_labels: true
show_conf: true
show_boxes: true
line_width: null
format: torchscript
keras: false
optimize: false
int8: false
dynamic: false
simplify: true
opset: null
workspace: null
nms: false
lr0: 0.01
lrf: 0.01
momentum: 0.937
weight_decay: 0.0005
warmup_epochs: 3.0
warmup_momentum: 0.8
warmup_bias_lr: 0.0
box: 0.1
cls: 0.1
dfl: 0.1
pose: 12.0
kobj: 1.0
nbs: 64
hsv_h: 0.015
hsv_s: 0.7
hsv_v: 0.4
degrees: 15.0
translate: 0.5
scale: 0.5
shear: 0.0
perspective: 0.0
flipud: 0.0
fliplr: 0.5
bgr: 0.0
mosaic: 1.0
mixup: 0.15
cutmix: 0.0
copy_paste: 0.3
copy_paste_mode: flip
auto_augment: randaugment
erasing: 0.0
cfg: null
tracker: botsort.yaml
After 500 epoches I have metrics:
Preccision - 0.934
Recall - 0.941
mAP50 - 0.973
mAP50-95 - 0.77
The accuracy on a real video is about 92%
I want to know:
Is this metrics already at its limit or can it be improved?
If can be, what arguments and how I should change?
Thank you for advance!