Filtern
Dokumenttyp
- Wissenschaftlicher Artikel (2)
- Preprint (1)
Schlagworte
- Deep Learning (3) (entfernen)
Institut
Advancements in Hand-Drawn Chemical Structure Recognition through an Enhanced DECIMER Architecture
(2024)
Accurate recognition of hand-drawn chemical structures is crucial for digitising hand-written chemical information found in traditional laboratory notebooks or for facilitating stylus-based structure entry on tablets or smartphones. However, the inherent variability in hand-drawn structures poses challenges for existing Optical Chemical Structure Recognition (OCSR) software. To address this, we present an enhanced Deep lEarning for Chemical ImagE Recognition (DECIMER) architecture that leverages a combination of Convolutional Neural Networks (CNNs) and Transformers to improve the recognition of hand-drawn chemical structures. The model incorporates an EfficientNetV2 CNN encoder that extracts features from hand-drawn images, followed by a Transformer decoder that converts the extracted features into Simplified Molecular Input Line Entry System (SMILES) strings. Our models were trained using synthetic hand-drawn images generated by RanDepict, a tool for depicting chemical structures with different style elements. To evaluate the model's performance, a benchmark was performed using a real-world dataset of hand-drawn chemical structures. The results indicate that our improved DECIMER architecture exhibits a significantly enhanced recognition accuracy compared to other approaches.
Advancements in hand-drawn chemical structure recognition through an enhanced DECIMER architecture
(2024)
Accurate recognition of hand-drawn chemical structures is crucial for digitising hand-written chemical information in traditional laboratory notebooks or facilitating stylus-based structure entry on tablets or smartphones. However, the inherent variability in hand-drawn structures poses challenges for existing Optical Chemical Structure Recognition (OCSR) software. To address this, we present an enhanced Deep lEarning for Chemical ImagE Recognition (DECIMER) architecture that leverages a combination of Convolutional Neural Networks (CNNs) and Transformers to improve the recognition of hand-drawn chemical structures. The model incorporates an EfficientNetV2 CNN encoder that extracts features from hand-drawn images, followed by a Transformer decoder that converts the extracted features into Simplified Molecular Input Line Entry System (SMILES) strings. Our models were trained using synthetic hand-drawn images generated by RanDepict, a tool for depicting chemical structures with different style elements. A benchmark was performed using a real-world dataset of hand-drawn chemical structures to evaluate the model's performance. The results indicate that our improved DECIMER architecture exhibits a significantly enhanced recognition accuracy compared to other approaches.
Fruits (follicles) of Hakea salicifolia and Hakea sericea (Proteaceae) are characterised by pronounced lignification and open via a ventral suture and the dorsal side. The opening along both sides is unique within the Proteaceae. Both serotinous species are obligate seeders, whose spreading benefits from bush fire events. The different tissues and the course of the vascular bundles must allow the opening mechanism. While their 2D-arrangements are known to some extent from light-microscopy images of cross-sections, this work presents their three-dimensional structures and discusses their contribution to the opening of Hakea fruits. For this purpose, 3D greyscale images, reconstructed from µCT-projection data of both fruits are segmented, assisted by a deep learning algorithm (AI algorithm). 3D renderings from these segmentations show strongly interconnected vascular bundles that build a double-dome shaped network in each valve of H. salicifolia and a dome shaped honeycomb-structure in each valve of H. sericea. However, the vascular bundles of both species show no interconnection between the two lateral valves of the fruit but leave gaps for predetermined fracture tissues on the ventral and dorsal side. The opening of the fruits after a fire or after separation from the mother plant can be explained by the anisotropic shrinkage in the two valves of the fruit.