Abstract
There are several new plant species that are discovered worldwide every year where many of these plants are toxic. In UAE there are many different plants that are considered toxic and may harm animals and humans. In this paper, we propose a deep-learning classification approach for plants in UAE. The aims of this study to determine the plants, trees or shrubs found in UAE either toxic or non-toxic. We train and test four different pre-trained convolutional neural network (CNN) models for image classification of seven classes of different types of plants. The CNN models are SqueezeNet, GoogleNet, DarkNet-53 and ResNet-50. Eeach of these networks shows different validation and testing accuracies and results. ResNet achieved the best performance with 97.94% validation accuracy AND 90.71% testing accuracy.
Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 3
Claim · model-written summary
ResNet classified toxic and non-toxic plants in UAE with 97.94% validation accuracy
- Taxon
- plantsViridiplantaee1
- Outcome
- validation accuracye3
- Value
- 97.94%e3
Verbatim
e1“we propose a deep-learning classification approach for plants in UAE”
e3“ResNet achieved the best performance with 97.94% validation accuracy”
Claim · model-written summary
ResNet classified toxic and non-toxic plants in UAE with 90.71% testing accuracy
- Taxon
- plantsViridiplantaee1
- Outcome
- testing accuracye4
- Value
- 90.71%e4
Verbatim
e1“we propose a deep-learning classification approach for plants in UAE”
e4“90.71% testing accuracy”
Claim · model-written summary
ResNet achieved the best performance among four pre-trained CNN models for plant image classification
Verbatim
e1“we propose a deep-learning classification approach for plants in UAE”
e2“The CNN models are SqueezeNet, GoogleNet, DarkNet-53 and ResNet-50”
e3“ResNet achieved the best performance with 97.94% validation accuracy”