Dataset for: Towards automatic smoke sensors inspection: Recognition of the smoke sensors in the industrial facilities and preparation for the future drone integration

This dataset stands for the paper: Towards automatic smoke sensors inspection: Recognition of the smoke sensors in the industrial facilities and preparation for the future drone integration

Dataset is saved in the SmokeSensorDetection.zip folder. 

The main directory contains two folders - Generated and Real. The first folder Generated contains data generated from renders of BOSCH and ESSER smoke sensors together with the Real photos of sensors inserted to the real captures of backgrounds. Each of these subfolders contains train and valid directories with the images and labels in YOLO and COCO format.

The second folder Real contains only photos of the real smoke sensors in the real environment. It contains the test, test_diff, train and valid subfolders, each with the images and labels in YOLO and COCO format. For testing, only test and test_diff directories were used.

It follows the following directory structure:

-- SmokeSensorDetection

   -- Generated

      -- BOSCH

         -- train

            -- images

            -- labels

            -- outputGeneratedBOSCHTrain.json

            -- classes.txt

         -- valid

            -- images

            -- labels

            -- outputGeneratedBOSCHVal.json

            -- classes.txt

      -- ESSER

         -- train

            -- images

            -- labels

            -- outputGeneratedESSERTrain.json

            -- classes.txt

         -- valid

            -- images

            -- labels

            -- outputGeneratedESSERVal.json

            -- classes.txt

      -- Real

         -- train

            -- images

            -- labels

            -- outputGeneratedRealTrain.json

         -- valid

            -- images

            -- labels

            -- outputGeneratedRealVal.json

   -- Real

         -- test

            -- images

            -- labels

            -- outputRealTest.json

         -- test_diff

            -- images

            -- labels

            -- outputRealTestDiff.json

         -- train

            -- images

            -- labels

            -- outputRealTrain.json

         -- valid

            -- images

            -- labels

            -- outputRealVal.json

Weights of the best performing models are saved in the ModelWeights.zip folder. 

The main directory contains three folders named by the model variant with the model checkpoints and weights.

It follows the following directory structure:

-- ModelWeights

   -- RT_DETRv2_mix_real_L

  -- SSD_real_real_VGG16

   -- YOLOv11_mix_real_N