Custom training parameters used for Fall_Detection Project

Hey Everyone!
I wanted to ask if anybody knows what training parameters were used for this specific model generation (mAP 88.6%, Precision 85.1%, Recall 86.1%):

I mean the training parameters given while starting the training like:

yolo task=detect mode=train data={dataset.location}/data.yaml epochs=25 imgsz=800 plots=True

Because when I follow the above for the same dataset, I get much lower values of mAP, Precision and Recall. Why is that so

Project Type: Object detection
Operating System & Browser: Windows and Google Chrome
Project Link: How to Use the Fall Detection Object Detection API

Hi @Muhammad_Nauman

It looks like the model in the link you sent was trained on Roboflow Train. You can learn how to use it in the link I attached.

Thanks! Is there any way to know the specific parameters only if I want to train it locally on my GPU?

Hi @Muhammad_Nauman

We don’t publish what specific parameters are used for training a specific model on Roboflow Train. From the dataset version that it was generated on, I see that it was resized to 640 x 640, but I don’t have more information beyond that.

Would you mind sharing what your use case for training the model locally would be?

I want to compare fall detection rates wrt age group for a survey

I have another query. Can the Trained Model API for this project be used directly with an rtsp camera?

We do have a variety of deployment methods available, but we don’t have something to directly connect to a RTSP camera at the moment.