GPU based Re-trainable Pruned CNN design for Camera Trapping at the Edge

Rohilla, Rajesh; Banga, Pavneet Singh; Garg, Pulkit; Mittal, Priya

doi:10.1109/icesc48915.2020.9155885

Cited by 1 publication

(3 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Edge devices like Raspberry Pi [45][46][47][48][49][50][51][52] and Nvidia Jetson [45,[53][54][55][56][57] have been previously used for edge-based detection of animals. Raspberry Pi 4 Model B is a single-board computer (SBC) with a Broadcom BCM2711 System-on-a-Chip (SoC) comprising of a Quadcore Cortex-A72 (ARM v8) 64-bit 1.5 GHz CPU and a VideoCore VI 3D 500 MHz GPU.…”

Section: Using Edge Devices For Camera Trap Animal Classificationmentioning

confidence: 99%

“…In addition to using the standardized optimization frameworks, custom optimizations can also be utilized. For example, a pruned CNN architecture that resulted in a smaller, less complex network that had the same performance as the original network was proposed by Rohilla et al [56]. In doing so, they were able to reduce the number of parameters of a five-layer CNN model by 94% (from 40,000 to 2400 parameters).…”

Section: Using Edge Devices For Camera Trap Animal Classificationmentioning

confidence: 99%

“…MobileNetV2 also provided a suitable model size of 40 MB. Rohilla et al [56] used the Nvidia Jetson TX1 to process images captured by camera traps. They trained a five-layer CNN on 27,370 fish images captured from a live video feed and pruned the trained model by removing all of the zero activation neurons.…”

Section: Using Edge Devices For Camera Trap Animal Classificationmentioning

confidence: 99%

See 2 more Smart Citations

An IoT System Using Deep Learning to Classify Camera Trap Images on the Edge

Zualkernan

Dhou

Judas³

et al. 2022

Computers

View full text Add to dashboard Cite

Camera traps deployed in remote locations provide an effective method for ecologists to monitor and study wildlife in a non-invasive way. However, current camera traps suffer from two problems. First, the images are manually classified and counted, which is expensive. Second, due to manual coding, the results are often stale by the time they get to the ecologists. Using the Internet of Things (IoT) combined with deep learning represents a good solution for both these problems, as the images can be classified automatically, and the results immediately made available to ecologists. This paper proposes an IoT architecture that uses deep learning on edge devices to convey animal classification results to a mobile app using the LoRaWAN low-power, wide-area network. The primary goal of the proposed approach is to reduce the cost of the wildlife monitoring process for ecologists, and to provide real-time animal sightings data from the camera traps in the field. Camera trap image data consisting of 66,400 images were used to train the InceptionV3, MobileNetV2, ResNet18, EfficientNetB1, DenseNet121, and Xception neural network models. While performance of the trained models was statistically different (Kruskal–Wallis: Accuracy H(5) = 22.34, p < 0.05; F1-score H(5) = 13.82, p = 0.0168), there was only a 3% difference in the F1-score between the worst (MobileNet V2) and the best model (Xception). Moreover, the models made similar errors (Adjusted Rand Index (ARI) > 0.88 and Adjusted Mutual Information (AMU) > 0.82). Subsequently, the best model, Xception (Accuracy = 96.1%; F1-score = 0.87; F1-Score = 0.97 with oversampling), was optimized and deployed on the Raspberry Pi, Google Coral, and Nvidia Jetson edge devices using both TenorFlow Lite and TensorRT frameworks. Optimizing the models to run on edge devices reduced the average macro F1-Score to 0.7, and adversely affected the minority classes, reducing their F1-score to as low as 0.18. Upon stress testing, by processing 1000 images consecutively, Jetson Nano, running a TensorRT model, outperformed others with a latency of 0.276 s/image (s.d. = 0.002) while consuming an average current of 1665.21 mA. Raspberry Pi consumed the least average current (838.99 mA) with a ten times worse latency of 2.83 s/image (s.d. = 0.036). Nano was the only reasonable option as an edge device because it could capture most animals whose maximum speeds were below 80 km/h, including goats, lions, ostriches, etc. While the proposed architecture is viable, unbalanced data remain a challenge and the results can potentially be improved by using object detection to reduce imbalances and by exploring semi-supervised learning.

show abstract