Warts

The main training dataset contains 25331 dermoscopic images, acquired at multiple sites and with different preprocessing methods applied beforehand. It contains images of the classes melanoma (MEL), melanocytic nevus (NV), basal cell carcinoma (BCC), actinic keratosis (AK), benign keratosis (BKL), dermatofibroma (DF), vascular lesion (VASC) and squamous cell carcinoma (SCC). A part of the training dataset is the HAM10000 dataset which contains images of size 600 × 450 that were centered and cropped around the lesion. The dataset curators applied histogram corrections to some images [1] . Another dataset, BCN_20000, contains images of size 1024 × 1024. This dataset is particularly challenging as many images are uncropped and lesions in difficult and uncommon locations are present [2] . Last, the MSK dataset contains images with various sizes.
The dataset also contains meta-information about the patient's age group (in steps of five years), the anatomical site (eight possible sites) and the sex (male/female). The meta data is partially incomplete, i.e., there are missing values for some images.
In addition, we make use of external data. We use the 955 dermoscopic images from the 7-point dataset [3] . Moreover, we use an in-house dataset which consists of 986 images. For the unknown class, we use 353 images obtained from a web search. We include images of healthy skin, angiomas, warts, cysts, and other benign alterations. The key idea is to build a broad class of skin variations that should encourage the model to assign any image that is not part of the eight main classes to the ninth broad pool of skin alterations. We also consider the three types of meta data for our external data, if it is available.
For internal evaluation, we split the main training dataset into five folds. The dataset contains multiple images of the same lesions. Thus, we ensure that all images of the same lesion are in the same fold. We add all our external data to each of the training sets. Note that we do not include any of our images from the unknown class in our evaluation as we do not know whether they accurately represent the actual unknown class. Thus, all our models are trained to predict nine classes but we only evaluate on the known, eight classes.
We use the mean sensitivity for our internal evaluation which is defined as $S=\frac{1}{C}\sum _{i=1}^C\frac{T{P}_i}{T{P}_i+F{N}_i}$ where TP are true positives, FN are false negatives and C is the number of classes. The metric is also used for the final challenge ranking.

Approximately, a total of 85 R. roxellanae were available for research in Shennongjia Nature Reserve. The monkeys were distributed across three regions: Region I: there were about 60 free-ranging monkeys lived in the Dalongtan area, which were composed of four one-male units (OMU) and one all male unit (AMU); Region II: Shennongjia snub-nosed monkey breeding base where 15 monkeys were kept and randomly divided into three OMUs by animal keepers; each unit was independently caged in one steel house; and Region III: Xiaolongtan area is the place where the injured monkeys rescued from different areas of Shennongjia Reserve were cared; 10 monkeys in a cage were taken as one unit. In total, we collected fresh fecal samples from 38 monkeys including 30 wild monkeys from Regions I and eight captive monkeys from II (six, three healthy and three diarrheal) and III (two, one healthy and one diarrheal monkeys) respectively in July 2012. Monkey age, raising patterns, sex, and health status were recorded (additional file 1). The monkey age was determined by their coat color, body length and wart-like growth at the corners of the mouth as described previously [27 ]. Meanwhile, the health status of the monkeys was clinically determined by observing their bright coats, movement (lively, powerful), good appetite, fecal shape, and color.
All the wild and captive monkeys were fed three times per day (10:00–11:00, 14:00–15:00, and 18:00–19:00). The wild monkeys from region I usually came quickly from the wild to the trees in the feeding place at the fixed time after hearing the call of the staff. Then they jumped down from the trees to get the food and ate it. After feeding, they immediately went back to the mountain for their free-living in the wild and could get more food in the forest. Meanwhile, the monkeys in regions II and III were kept and fed in their cages at similar timepoints.
Fecal samples were collected from the monkeys by using the protocol described previously [2 (link)]. Briefly, fecal samples were collected during one week by two researchers before feeding time in the morning (10:00–11:00) and afternoon (14:00–15:00), when the monkeys had already gone down the mountain, gathered in the trees at the feeding place and waited for their food. Each researcher was responsible for one monkey unit per time. Fresh fecal samples were immediately collected into sterile tubes. For the diarrheal fecal samples, sterile cotton swabs were used and dipped the feces in the center carefully for several times to collect feces as many as possible, and stored into sterile tubes as well. Samples were stored at -80 °C at Shennongjia Reserve and transported to our laboratory on dry ice and stored at -80 °C until DNA extraction. Microbial DNA from fecal samples was extracted using QIAamp DNA stool mini kits (Qiagen, CA, USA) following standard protocols.