VLP Dataset Metadata

library_name: keras tags: - Clasificacion imagen-detection - Deteccion LED

Model description

This dataset was acquired during the dissertation entitled Optical Camera Communications and Machine Learning for Indoor Visible Light Positioning. This work was carried out in the academic year 2020/2021 at the Instituto de Telecomunicacoes in Aveiro.

The images that constitute this dataset were acquired over a grid with 15 regularly spaced reference points on the floor surface. Table 2 shows the position of these points in relation to the referential defined in the room along with the position of the LED luminaires. During the dataset acquisition, the CMOS image sensor (Sony IMX219) was positioned parallel to the floor at a height of 25.6 cm facing upwards, i.e. with pitch and yaw angles equal to 0. All images were saved as TIFF (Tagged Image File Format) with a resolution of 3264 × 2464 pixels and exposure and readout times equal to 9 µs and 18 µs, respectively.

Intended uses & limitations

Can use as academic and formation objetives

Training and evaluation data

Training by a Keras model

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

Hyperparameters	Value
name	RMSprop
learning_rate	0.0010000000474974513
decay	0.0
rho	0.8999999761581421
momentum	0.0
epsilon	1e-07
centered	False
training_precision	float32

Model Plot

View Model Plot