Lightweight bark image network accurately identifies Bangladeshi trees
BarkNet-Lite: A Lightweight Texture and Colour Network with the BarkBD Benchmark for Bark-Based Tree Species Recognition in Bangladesh
Computer Vision and Pattern Recognition
Summary
Identifying tree species helps track forests and biodiversity but usually needs expert knowledge. The authors created a new picture collection of tree bark from Bangladesh and built a small, fast computer model that learns from scratch how to recognize these tree types from bark images. Their model works almost as well as big models trained on large generic datasets but runs efficiently on a smartphone. Tests show it focuses on bark details, not the background, making its decisions reliable.
What this means in practice
- •For forest inventory teams: Use a compact smartphone model to identify tree species from bark images in Bangladesh, enabling faster and expert-independent forest surveys.
- •For mobile app developers: Integrate a lightweight bark recognition model into smartphone apps to classify tree species quickly without relying on cloud computing.$Commercial implications: Enables selling user-friendly tree identification apps for environmental monitoring and education using an efficient on-device model.
Authors
Aroshi Ali, Saad Ahmed, Md. Khalid Syfullah
Abstract
Tree species recognition supports forest inventory and biodiversity monitoring but still depends on scarce taxonomic expertise. Bark is visible year-round at ground level, yet bark recognition has concentrated on temperate floras and on large ImageNet-pre-trained backbones. We address both gaps. First, we release BarkBD, a bark dataset for Bangladesh: 14,258 uncropped smartphone photographs of 20 native species across four districts and three weather conditions, with a fixed stratified split. Second, we propose BarkNet-Lite, a 2.96M-parameter network trained from random initialisation, pairing a multi-scale texture pathway with a parallel colour-aware pathway. Over five seeds it reaches 96.64+-0.66%accuracyunderstrict single-image inference, within 2.3 points of nine ImageNet-pre-trained backbones fine-tuned under an identical protocol and within one seed-level standard deviation of the smallest ofthem, andtransfers to public benchmarks (95.86% on BarkVN-50, 92.85% on BarkNet 1.0). Grad-CAM, validated by faithfulness and weight-randomisation checks, confirms its decisions rest on bark structure rather than background. The exported single-precision model classifies one photograph in 15.34ms on a commodity smartphone.