Transfer Learning by Finetuning Pretrained CNNs Entirely with Synthetic Images

Rajpura, Param S.; Aggarwal, Alakh; Goyal, Manik; Gupta, Sanchit; Talukdar, Jonti; Bojinov, Hristo; Hegde, Ravi S.

doi:10.1007/978-981-13-0020-2_45

Cited by 14 publications

(10 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The process of fine-tuning pre-trained models with an expanded database has precedent in existing research works [25]. It can be compared with the closely related field of transfer learning [26].…”

Section: E Fine Tuning Of Pre-trained Modelsmentioning

confidence: 99%

Adapting Data-Driven Techniques to Improve Surrogate Machine Learning Model Performance

Jones

Popescu

et al. 2023

IEEE Access

View full text Add to dashboard Cite

We demonstrate the adaption of three established methods to the field of surrogate machine learning model development. These methods are data augmentation, custom loss functions and fine-tuning of pre-trained models. Each of these methods have seen widespread use in the field of machine learning, however, here we apply them specifically to surrogate machine learning model development. The machine learning model that forms the basis behind this work was intended to surrogate a traditional engineering model used in the UK nuclear industry. Previous performance of this model has been hampered by poor performance due to limited training data. Here, we demonstrate that through a combination of additional techniques, model performance can be significantly improved. We show that each of the aforementioned techniques have utility in their own right and in combination with one another. However, we see them best applied when used to fine-tune existing models. Five pre-trained surrogate models produced prior to this study were further trained using an augmented dataset and with our custom loss function. Through the combination of all three techniques, we see an improvement of at least 38% in performance across the five models.

show abstract

Section: E Fine Tuning Of Pre-trained Modelsmentioning

confidence: 99%

Adapting Data-Driven Techniques to Improve Surrogate Machine Learning Model Performance

Jones

Popescu

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The idea of DR has been applied to bridging reality gap in many tasks from car detection (Tremblay et al, 2018) to detecting packaged food in refrigerators (Rajpura et al, 2017). Specific to image-based plant phenotyping, applied DR to leaf instance segmentation.…”

Section: Related Workmentioning

confidence: 99%

Scalable learning for bridging the species gap in image-based plant phenotyping

Ward¹,

Moghadam²

2020

Preprint

View full text Add to dashboard Cite

The traditional paradigm of applying deep learning -collect, annotate and train on data-is not applicable to image-based plant phenotyping. Data collection involves the growth of many physical samples, imaging them at multiple growth stages and finally manually annotating each image. This process is error-prone, expensive, time consuming and often requires specialised equipment. Almost 400,000 different plant species exist across the world. Each varying greatly in appearance, geometry and structure, a species gap exists between the domain of each plant species. The performance of a model is not generalisable and may not transfer to images of an unseen plant species. With the costs of data collection and number of plant species, it is not tractable to apply deep learning to the automation of plant phenotyping measurements. Hence, training using synthetic data is effective as the cost of data collection and annotation is free. We investigate the use of synthetic data for image-based plant phenotyping. Our conclusions and released data are applicable to the measurement of phenotypic traits including plant area, leaf count, leaf area and shape. In this paper, we validate our proposed approach on leaf instance segmentation for the measurement of leaf area. We study multiple synthetic data training regimes using Mask-RCNN when few or no annotated real data is available. We also present UPGen: a Universal Plant Generator for bridging the species gap. UPGen leverages domain randomisation to produce widely distributed data samples and models stochastic biological variation. A model trained on our synthetic dataset traverses the domain and species gaps. In validating UPGen, the relationship between different data parameters and their effects on leaf segmentation performance is investigated. Imitating a plant phenotyping facility processing a new plant species, our methods outperform standard practices, such as transfer learning from publicly available plant data, by 26.6% and 51.46% on two unseen plant species respectively. We benchmark UPGen by using it to compete in the CVPPP Leaf Segmentation Challenge. Generalising across multiple plant species, our method achieved state-of-the-art performance scoring a mean of 88% across A1-4 test datasets. Our synthetic dataset and pretrained model are available at https://danielcward.github.io/UPGen/.

show abstract

“…Object detection or more precisely one of its multiple kinds is probably the most pursued topic of recent research in the area of training an artificial intelligence on synthetic data. Works in this direction includes Rajpura et al 19 use Blender and Cyles to generate synthetic image data for the purpose of object detection by transfer learning online 3D models from ShapeNet database (with everyday items like bottles, tins, cans and food items) and Archive3D database. They also encounter domain gap problem Jabbar et al conducted experiments to detect drinking glasses in realistically rendered images compared to real images.…”

Section: Object Detectionmentioning

confidence: 99%

Physically based synthetic image generation for machine learning: a review of pertinent literature

Schraml

2019

Photonics and Education in Measurement Science 2019

View full text Add to dashboard Cite

The term deep learning is almost on everyone's lips these days, in the area of computer vision mainly because of the great advances deep learning approaches have made amongst others in object detection and classification. For general object localization or classification tasks there do exist several giant databases containing several millions of labeled images and several thousands of different labels like COCO and ImageNet. In contrast in industrial applications like quality inspection there do hardly ever exist such training data not only for reasons of confidentiality of trade secrets. An obvious way to remedy this deficiency is the synthetic creation of image data. Physically based rendering attempts to achieve photorealistic images by accurately simulating the flow of light of the real world according to various physical laws. Therefor multiple techniques like Ray Tracing and Path Tracing have been implemented and are becoming increasingly widespread as hardware performance increases. The intent of this article is to give a wide but nevertheless preferably comprehensive overview of which approaches have been pursued in recent literature to generate realistic synthetic training images. The development of various rendering methods from rasterization to bidirectional monte carlo path tracing is outlined, as well as their differences and use. Along with the terminology a few mathematical foundations like the Bidirectional Reflectance Distribution Function (BRDF) are briefly described. Altogether special concern is given to industrial data and quality control, comparing literature and the practical application of its results.

show abstract

Transfer Learning by Finetuning Pretrained CNNs Entirely with Synthetic Images

Cited by 14 publications

References 23 publications

Adapting Data-Driven Techniques to Improve Surrogate Machine Learning Model Performance

Adapting Data-Driven Techniques to Improve Surrogate Machine Learning Model Performance

Scalable learning for bridging the species gap in image-based plant phenotyping

Physically based synthetic image generation for machine learning: a review of pertinent literature

Contact Info

Product

Resources

About