config.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# File: config.py

import numpy as np

# for new data
NORM = 'BN'
PAN = False
PSEUDO = False
CASCADE = True
SOFTNMS = True
RESNET = False
RESNET_BATCH = 32
RESNET_SIZE = 384
SAMPLING = True
RELATION = False
OHEM = True
FREEZE_C2 = True
MR = False
FOCAL = False
# mode flags ---------------------
MODE_MASK = True
FPN = True if CASCADE else False
FPN_STRIDES = [64, 32, 16, 8, 4]
FPN_SIZES = [512, 256, 128, 64, 32]
#FPN_SIZES = [256, 128, 64, 32, 16]
# dataset -----------------------
BASEDIR = '/data/dataset/airbus/'
TRAIN_DATASET = ['train_v2'] if not RESNET else ['train']
if PSEUDO:
    TRAIN_DATASET=['train_v2', 'test']
VAL_DATASET = 'val_v2' if not RESNET else ['val']
TEST_DATASET = 'test'
NUM_CLASS = 2
 # NUM_CLASS strings
CLASS_NAMES = ['BG', 'SHIP']


# basemodel ----------------------
#RESNET_NUM_BLOCK = [3, 4, 6, 3]     # resnet50
RESNET_NUM_BLOCK = [3, 4, 23, 3]     # resnet101

# preprocessing --------------------
SHORT_EDGE_SIZE = 2000
MAX_SIZE = 2500
#SHORT_EDGE_SIZE = 200
#MAX_SIZE = 200
# alternative (better) setting: 800, 1333

# anchors -------------------------
ANCHOR_STRIDE = 16
#ANCHOR_STRIDE = 4
# sqrtarea of the anchor box
ANCHOR_SIZES = (32, 64, 128, 256, 512)
#ANCHOR_SIZES = (8, 16, 32, 64, 128)
ANCHOR_RATIOS = (0.5, 1., 2.)
NUM_ANCHOR = len(ANCHOR_SIZES) * len(ANCHOR_RATIOS)
POSITIVE_ANCHOR_THRES = 0.7
NEGATIVE_ANCHOR_THRES = 0.3
# just to avoid too large numbers.
BBOX_DECODE_CLIP = np.log(MAX_SIZE / 16.0)

# rpn training -------------------------
# keep fg ratio in a batch in this range
RPN_FG_RATIO = 0.5
RPN_BATCH_PER_IM = 256
RPN_MIN_SIZE = 0
RPN_PROPOSAL_NMS_THRESH = 0.7
TRAIN_PRE_NMS_TOPK = 12000
TRAIN_POST_NMS_TOPK = 2000
# boxes overlapping crowd will be ignored.
CROWD_OVERLAP_THRES = 0.7

# fastrcnn training ---------------------
FASTRCNN_BATCH_PER_IM = 256
FASTRCNN_BBOX_REG_WEIGHTS = np.array([10, 10, 5, 5], dtype='float32')
FASTRCNN_FG_THRESH = 0.5
# keep fg ratio in a batch in this range
FASTRCNN_FG_RATIO = 0.25

# testing -----------------------
TEST_PRE_NMS_TOPK = 6000
TEST_POST_NMS_TOPK = 600
FASTRCNN_NMS_THRESH = 0.5
RESULT_SCORE_THRESH = 0.5
RESULTS_PER_IM = 100