utils.py

import numpy as np
import torch
import pandas as pd
import random
import string

# https://github.com/benhamner/Metrics/blob/master/Python/ml_metrics/average_precision.py
def apk(actual, predicted, k=10):
    if len(predicted)>k:
        predicted = predicted[:k]

    score = 0.0
    num_hits = 0.0

    for i,p in enumerate(predicted):
        if p in actual and p not in predicted[:i]:
            num_hits += 1.0
            score += num_hits / (i+1.0)

    if not actual:
        return 0.0

    return score / min(len(actual), k)

def mapk(actual, predicted, k=10):
    return np.mean([apk(a,p,k) for a,p in zip(actual, predicted)])

def map5kfast(preds, targs, k=10):
    predicted_idxs = preds.sort(descending=True)[1]
    top_5 = predicted_idxs[:, :5]
    scores = torch.zeros(len(preds), k).float()
    for kk in range(k):
        scores[:,kk] = (top_5[:,kk] == targs).float() / float((kk+1))
    return scores.max(dim=1)[0].mean()

def map5(preds,targs):
    if type(preds) is list:
        return torch.cat([map5fast(p, targs, 5).view(1) for p in preds ]).mean()
    return map5kfast(preds,targs, 5)

def top_5_preds(preds): return np.argsort(preds.numpy())[:, ::-1][:, :5]

def top_5_pred_labels(preds, classes):
    top_5 = top_5_preds(preds)
    labels = []
    for i in range(top_5.shape[0]):
        labels.append(' '.join([classes[idx] for idx in top_5[i]]))
    return labels

def create_submission(preds, data, name, classes=None):
    if not classes: classes = data.classes
    sub = pd.DataFrame({'Image': [path.name for path in data.test_ds.x.items]})
    sub['Id'] = top_5_pred_labels(preds, classes)
    sub.to_csv(f'subs/{name}.csv.gz', index=False, compression='gzip')


def intersection(preds, targs):
    # preds and targs are of shape (bs, 4), pascal_voc format
    max_xy = torch.min(preds[:, 2:], targs[:, 2:])
    min_xy = torch.max(preds[:, :2], targs[:, :2])
    inter = torch.clamp((max_xy - min_xy), min=0)
    return inter[:, 0] * inter[:, 1]

def area(boxes):
    return ((boxes[:, 2]-boxes[:, 0]) * (boxes[:, 3]-boxes[:, 1]))

def union(preds, targs):
    return area(preds) + area(targs) - intersection(preds, targs)

def IoU(preds, targs):
    return intersection(preds, targs) / union(preds, targs)

def name(n=10, print_it=True):
    name = "".join(random.choice(string.ascii_lowercase) for _ in range(n))
    if print_it: print(name)
    return name