No abstract

In this paper we challenge the common assumption that convolutional layers in modern CNNs are translation invariant. We show that CNNs can and will exploit the absolute spatial location by learning filters that respond exclusively to particular absolute locations by exploiting image boundary effects. Because modern CNNs filters have a huge receptive field, these boundary effects operate even far from the image boundary, allowing the network to exploit absolute spatial location all over the image. We give a simple solution to remove spatial location encoding which improves translation invariance and thus gives a stronger visual inductive bias which particularly benefits small data sets. We broadly demonstrate these benefits on several architectures and various applications such as image classification, patch matching, and two video classification datasets.

A novel automatic Rapid Diagnostic Test (RDT) reader platform is designed to analyze and diagnose target disease by using existing consumer cameras of a laptop-computer or a tablet. The RDT reader is useable with numerous lateral immunochromatographic assays and similar biomedical tests. The system has two different components, which are 3D-printed, low-cost, tiny, and compact stand and a decision program named RDT-AutoReader 2.0. The program takes the image of RDT, crops the region of interest (ROI), and extracts the features from the control end test lines to classify the results as invalid, positive, or negative. All related patient's personal information, image of ROI, and the e-report are digitally saved and transferred to the related clinician. Condition of the patient and the progress of the disease can be monitored by using the saved data. The reader platform has been tested by taking image from used cassette RDTs of rotavirus (RtV)/adenovirus (AdV) and lateral flow strip RDTs of Helicobacter pylori (H. pylori) before discarding them. The created RDT reader can also supply real-time statistics of various illnesses by using databases and Internet. This can help to inhibit propagation of contagious diseases and to increase readiness against epidemic diseases worldwide.

We show that object detectors can hallucinate and detect missing objects; potentially even accurately localized at their expected, but non-existing, position. This is particularly problematic for applications that rely on visual part verification: detecting if an object part is present or absent. We show how popular object detectors hallucinate objects in a visual part verification task and introduce the first visual part verification dataset: DelftBikes 1 , which has 10,000 bike photographs, with 22 densely annotated parts per image, where some parts may be missing. We explicitly annotated an extra object state label for each part to reflect if a part is missing or intact. We propose to evaluate visual part verification by relying on recall and compare popular object detectors on DelftBikes.