Datasets:

amir-salar
/

test-dataset

image image	modality string	label string
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both
	cv	both

End of preview. Expand in Data Studio

FASDD fire/smoke classification — test split

Held-out test split of the FASDD-derived (TDML) fire/smoke classification data, packaged for inference and accuracy checks of exported ONNX / TensorRT models (e.g. on a Jetson).

20,435 images across three modalities. Labels: fire, smoke, both, neither.

Modality	Folder	Images	Format	fire	smoke	both	neither
CV (ground-level)	`cv/`	15,884	JPEG	2,091	3,902	3,358	6,533
UAV	`uav/`	4,181	JPEG	35	846	1,303	1,997
Remote sensing	`rs/`	370	GeoTIFF/TIFF	0	31	191	148

Notes: UAV has very few fire-only images and RS has none, so per-class metrics for those cells are noisy or undefined — report per modality.

Files

metadata.csv has one row per image:

file_name — path relative to the repo root, laid out as <modality>/<label>/<file> (e.g. cv/smoke/smoke_CV022310.jpg)
modality — cv, uav or rs
label — one of fire, smoke, both, neither

Usage

git lfs install
git clone https://huggingface.co/datasets/amir-salar/test-dataset

import pandas as pd
df = pd.read_csv("test-dataset/metadata.csv")

License / attribution

Images originate from the FASDD dataset (Fire and Smoke Detection Dataset). Refer to the original FASDD license and citation terms for redistribution and usage conditions.

Downloads last month: -