Preservation of High Frequency Content for Deep Learning-Based Medical Image Classification

McIntosh, Declan; Marques, Tunai Porto; Albu, Alexandra Branzan

doi:10.1109/crv52889.2021.00010

Cited by 2 publications

(8 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Due to the growing domains where large-scale video collections are taking place, such as surveillance and marine environmental monitoring, there has been significant research interest in efficient methods of classifying video and video clips [16], [19]. Methods like those presented by Bhardwaj et al [20] focus on reducing the number of frames necessary for video classification.…”

Section: Related Workmentioning

confidence: 99%

“…1) A Spatio-Temporal Pooling Bridge where both temporal and spatial pooling are used together, decreasing model depth and improving efficiency, reference Sec III-A. 2) Residual blocks to improve training of our relatively deep network (two sequential encoders), see Sec III-B [24] 3) A temporal attention module during the spatial encoding stage of our model for adaptive feature refinement of the temporal axis before the temporal encoding stage, as in Sec III-C. 4) An energy-preserving step [16] for image downsampling which increases the number of input channels but preserves all of its original high frequency information (otherwise partially lost), as detailed in III-D.…”

Section: Proposed Approachmentioning

confidence: 99%

“…Gaussian filtering followed by subsampling) that eliminate high-frequency components of the input. In order to achieve the necessary resizing without neglecting high-frequency components, DWTs have been recently employed [16] before the first layers of CNNs. Studies show that this strategy can significantly increase the performance of CNN-based tasks where higher resolution images are available [16].…”

Section: Energy Preserving Wavelet Down-samplingmentioning

confidence: 99%

“…These wavelets create multiple representations of the input that carry all its information and coherently represent the information through the separation of frequency components. These additional input channels increase the number of input channels by a factor of 4 but only increase the number of model parameters and FLOPS by less than 1% [16], see Table I. This small cost provides the network with the information of an image with 4 times higher spatial resolution than otherwise.…”

Section: Energy Preserving Wavelet Down-samplingmentioning

confidence: 99%

“…=We propose an efficient novel deep learning-based event detection system, TempNet, which operates on 4-second video clips to detect organism behaviour. Our detector utilizes temporal attention modules based on previous channel attention modules, residual convolutional blocks, wavelet downsampling, and a custom architecture to improve performance over previous methods while achieving faster than real-time efficiency [15], [16]. TempNet not only outperforms the baseline method proposed in [14], but since it does not rely on species-specific visual features, it can be applied to any other behavioural event for which sufficient annotations have been compiled.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

TempNet: Temporal Attention Towards the Detection of Animal Behaviour in Videos

McIntosh

Marques

Albu

et al. 2022

2022 26th International Conference on Pattern Recognition (ICPR)

Self Cite

View full text Add to dashboard Cite

Recent advancements in cabled ocean observatories have increased the quality and prevalence of underwater videos; this data enables the extraction of high-level biologically relevant information such as species' behaviours. Despite this increase in capability, most modern methods for the automatic interpretation of underwater videos focus only on the detection and counting organisms. We propose an efficient computer vision-and deep learning-based method for the detection of biological behaviours in videos. TempNet uses an encoder bridge and residual blocks to maintain model performance with a two-staged, spatial, then temporal, encoder. TempNet also presents temporal attention during spatial encoding as well as Wavelet Down-Sampling preprocessing to improve model accuracy. Although our system is designed for applications to diverse fish behaviours (i.e, is generic), we demonstrate its application to the detection of sablefish (Anoplopoma fimbria) startle events. We compare the proposed approach with a state-of-the-art end-to-end video detection method (ReMotENet) and a hybrid method previously offered exclusively for the detection of sablefish's startle events in videos from an existing dataset. Results show that our novel method comfortably outperforms the comparison baselines in multiple metrics, reaching a per-clip accuracy and precision of 80% and 0.81, respectively. This represents a relative improvement of 31% in accuracy and 27% in precision over the compared methods using this dataset. Our computational pipeline is also highly efficient, as it can process each 4-second video clip in only 38ms. Furthermore, since it does not employ features specific to sablefish startle events, our system can be easily extended to other behaviours in future works.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Proposed Approachmentioning

confidence: 99%

Section: Energy Preserving Wavelet Down-samplingmentioning

confidence: 99%

Section: Energy Preserving Wavelet Down-samplingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

TempNet: Temporal Attention Towards the Detection of Animal Behaviour in Videos

McIntosh

Marques

Albu

et al. 2022

2022 26th International Conference on Pattern Recognition (ICPR)

Self Cite

View full text Add to dashboard Cite

show abstract

The intriguing effect of frequency disentangled learning on medical image segmentation

Fu,

Jimenez,

Loizillon

et al. 2024

Medical Imaging 2024: Image Processing

View full text Add to dashboard Cite

Deep models have been shown to tend to fit the target function from low to high frequencies (a phenomenon called the frequency principle of deep learning). One may hypothesize that such property can be leveraged for better training of deep learning models, in particular for segmentation tasks where annotated datasets are often small. In this paper, we exploit this property to propose a new training method based on frequencydomain disentanglement. It consists of three main stages. First, it disentangles the image into high-and low-frequency components. Then, the segmentation network model learns them separately (the approach is general and can use any segmentation network as backbone). Finally, feature fusion is performed to complete the downstream task. The method was applied to the segmentation of the red and dentate nuclei in Quantitative Susceptibility Mapping (QSM) data and to three tasks of the Medical Segmentation Decathlon (MSD) challenge under different training sample sizes. For segmenting the red and dentate nuclei and the heart, the proposed approach resulted in considerable improvements over the baseline (respectively between 8 and 16 points of Dice and between 5 and 8 points). On the other hand, there was no improvement for the spleen and the hippocampus. We believe that these intriguing results, which echo theoretical work on the frequency principle of deep learning, are of interest for discussion at the conference. The source code is publicly available at: https://github.com/GuanghuiFU/frequency_disentangled_learning.

show abstract

Preservation of High Frequency Content for Deep Learning-Based Medical Image Classification

Cited by 2 publications

References 22 publications

TempNet: Temporal Attention Towards the Detection of Animal Behaviour in Videos

TempNet: Temporal Attention Towards the Detection of Animal Behaviour in Videos

The intriguing effect of frequency disentangled learning on medical image segmentation

Contact Info

Product

Resources

About