arthurdouillard

class SelfAttention(nn.Module):
    def __init__(self, dim, num_heads=8):
        super().__init__()
        self.Wq = nn.Linear(dim, num_heads * dim, bias=False)
        self.Wk = nn.Linear(dim, num_heads * dim, bias=False)
        self.Wv = nn.Linear(dim, num_heads * dim, bias=False)
        self.Wo = nn.Linear(num_heads * dim, dim)

    def forward(self, x):
        # X is of shape (Batch size, number of tokens, embedding dimension)

arthurdouillard

# import the necessary packages
import numpy as np
 
# Malisiewicz et al.
def non_max_suppression_fast(boxes, overlapThresh):
	# if there are no boxes, return an empty list
	if len(boxes) == 0:
		return []
 
	# if the bounding boxes integers, convert them to floats --

arthurdouillard

def compute_metrics(true_pos, false_pos, false_neg):
    """Compute the precision, recall, and f1 score."""
    precision = true_pos / (true_pos + false_pos)
    recall = true_pos / (true_pos + false_neg)
    
    if precision == 0 or recall == 0:
        return precision, recall, 0
    
    f1 = 2 / (1 / precision + 1 / recall)
    return precision, recall, f1

arthurdouillard

def load_retinanet(weights, n_classes, freeze=True):
    modifier = freeze_model if freeze else None

    model = resnet50_retinanet(num_classes=num_classes, modifier=modifier)
    model.load_weights(weights, by_name=True, skip_mismatch=True)

    return model
  
  
def compile(model):

arthurdouillard

class DfGenerator(CSVGenerator):
    """Custom generator intented to work with in-memory Pandas' dataframe."""
    def __init__(self, df, class_mapping, cols, base_dir='', **kwargs):
        """Initialization method.

        Arguments:
            df: Pandas DataFrame containing paths, labels, and bounding boxes.
            class_mapping: Dict mapping label_str to id_int.
            cols: Dict Mapping 'col_{filename/label/x1/y1/x2/y2} to corresponding df col.
        """

arthurdouillard

def jaccard(box_a, box_b):
    """box_X is a tuple of the shape (x1, y1, x2, y2)."""
    side1 = max(0, min(a[2], b[2]) - max(a[0], b[0]))
    side2 = max(0, min(a[3], b[3]) - max(a[1], b[1]))
    inter = side1 * side2
    
    area_a = (a[2] - a[0]) * (a[3] - a[1])
    area_b = (b[2] - b[0]) * (b[3] - b[1])
    union = area_a + area_b - inter
    

arthurdouillard

from IPython.display import clear_output, Image, display, HTML

def strip_consts(graph_def, max_const_size=32):
    """Strip large constant values from graph_def."""
    strip_def = tf.GraphDef()
    for n0 in graph_def.node:
        n = strip_def.node.add() 
        n.MergeFrom(n0)
        if n.op == 'Const':
            tensor = n.attr['value'].tensor