to3i
908 linhas
32 KiB
Plaintext
908 linhas
32 KiB
Plaintext
{
|
|
"cells": [
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"## Test and visualize sign detector"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 1,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"import numpy as np\n",
|
|
"import pandas as pd\n",
|
|
"from PIL import Image\n",
|
|
"from ast import literal_eval\n",
|
|
"\n",
|
|
"from tqdm import tqdm\n",
|
|
"\n",
|
|
"import torch"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 2,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"import matplotlib.pyplot as plt"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 3,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"# for auto-reloading external modules\n",
|
|
"# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n",
|
|
"%load_ext autoreload\n",
|
|
"%autoreload 2"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 4,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"relative_path = '../../'\n",
|
|
"# ensure that parent path is on the python path in order to have all packages available\n",
|
|
"import sys, os\n",
|
|
"parent_path = os.path.join(os.getcwd(), relative_path)\n",
|
|
"parent_path = os.path.realpath(parent_path) # os.path.abspath(...)\n",
|
|
"sys.path.insert(0, parent_path)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 5,
|
|
"metadata": {
|
|
"scrolled": true
|
|
},
|
|
"outputs": [
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"/home/tobias/.virtualenvs/pytorch/local/lib/python2.7/site-packages/cryptography/hazmat/primitives/constant_time.py:26: CryptographyDeprecationWarning: Support for your Python version is deprecated. The next version of cryptography will remove support. Please upgrade to a 2.7.x release that supports hmac.compare_digest as soon as possible.\n",
|
|
" utils.DeprecatedIn23,\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"from lib.datasets.cunei_dataset_segments import CuneiformSegments, get_segment_meta\n",
|
|
"from lib.models.trained_model_loader import get_fpn_ssd_net\n",
|
|
"from lib.detection.run_gen_ssd_detection import gen_ssd_detections"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 6,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"collections = ['test'] # e.g. train test saa06\n",
|
|
"only_annotated = True\n",
|
|
"only_assigned = True\n",
|
|
"\n",
|
|
"# store detections for re-use\n",
|
|
"save_detections = False\n",
|
|
"\n",
|
|
"# show detections\n",
|
|
"show_detections = True"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 7,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"model_version = 'v191ft01' \n",
|
|
"\n",
|
|
"arch_type = 'mobile' # resnet, mobile\n",
|
|
"arch_opt = 1\n",
|
|
"width_mult = 0.625 # 0.5 0.625 0.75\n",
|
|
"\n",
|
|
"crop_shape = [600, 600]\n",
|
|
"tile_shape = [600, 600]\n",
|
|
"\n",
|
|
"num_classes = 240\n",
|
|
"\n",
|
|
"with_64 = False \n",
|
|
"create_bg_class = False \n",
|
|
"with_4_aspects = False "
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"<a id='complconf'>Config Completeness</a>\n",
|
|
"\n",
|
|
"[Jump to results](#results)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 8,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"test_min_score_thresh = 0.01 # 0.01 0.05\n",
|
|
"test_nms_thresh = 0.5 \n",
|
|
"\n",
|
|
"eval_ovthresh = 0.5 # 0.4"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 9,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"### Load Model"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 10,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"device = 'cuda' if torch.cuda.is_available() else 'cpu'"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 11,
|
|
"metadata": {
|
|
"scrolled": true
|
|
},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"FPNSSD(\n",
|
|
" (fpn): MobileNetV2FPN(\n",
|
|
" (features): Sequential(\n",
|
|
" (0): Sequential(\n",
|
|
" (0): Conv2d(1, 20, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" )\n",
|
|
" (1): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(20, 20, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(20, 20, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=20, bias=False)\n",
|
|
" (4): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(20, 10, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (2): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(10, 60, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(60, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(60, 60, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=60, bias=False)\n",
|
|
" (4): BatchNorm2d(60, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(60, 15, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(15, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (1): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(15, 90, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(90, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(90, 90, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=90, bias=False)\n",
|
|
" (4): BatchNorm2d(90, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(90, 15, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(15, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (3): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(15, 90, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(90, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(90, 90, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=90, bias=False)\n",
|
|
" (4): BatchNorm2d(90, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(90, 20, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (1): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(20, 120, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(120, 120, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=120, bias=False)\n",
|
|
" (4): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(120, 20, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (2): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(20, 120, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(120, 120, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=120, bias=False)\n",
|
|
" (4): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(120, 20, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(20, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (4): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(20, 120, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(120, 120, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), groups=120, bias=False)\n",
|
|
" (4): BatchNorm2d(120, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(120, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(40, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (1): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(240, 240, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=240, bias=False)\n",
|
|
" (4): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(40, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (2): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(240, 240, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=240, bias=False)\n",
|
|
" (4): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(40, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (3): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(240, 240, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=240, bias=False)\n",
|
|
" (4): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(240, 40, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(40, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (5): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(40, 240, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(240, 240, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=240, bias=False)\n",
|
|
" (4): BatchNorm2d(240, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(240, 60, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(60, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (1): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(60, 360, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(360, 360, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=360, bias=False)\n",
|
|
" (4): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(360, 60, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(60, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (2): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(60, 360, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(360, 360, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=360, bias=False)\n",
|
|
" (4): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(360, 60, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(60, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (6): MobileBlock(\n",
|
|
" (mobile_block): Sequential(\n",
|
|
" (0): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(60, 360, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(360, 360, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=360, bias=False)\n",
|
|
" (4): BatchNorm2d(360, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(360, 100, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (1): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(100, 600, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(600, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(600, 600, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=600, bias=False)\n",
|
|
" (4): BatchNorm2d(600, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(600, 100, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (2): InvertedResidual(\n",
|
|
" (conv): Sequential(\n",
|
|
" (0): Conv2d(100, 600, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(600, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" (3): Conv2d(600, 600, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=600, bias=False)\n",
|
|
" (4): BatchNorm2d(600, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (5): ReLU6(inplace)\n",
|
|
" (6): Conv2d(600, 100, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (7): BatchNorm2d(100, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" )\n",
|
|
" (7): Sequential(\n",
|
|
" (0): Conv2d(100, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
|
|
" (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
|
|
" (2): ReLU6(inplace)\n",
|
|
" )\n",
|
|
" (8): AvgPool2d(kernel_size=7, stride=1, padding=0)\n",
|
|
" )\n",
|
|
" (conv6): Conv2d(512, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1))\n",
|
|
" (toplayer): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))\n",
|
|
" )\n",
|
|
" (loc_head): Sequential(\n",
|
|
" (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (1): ReLU(inplace)\n",
|
|
" (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (3): ReLU(inplace)\n",
|
|
" (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (5): ReLU(inplace)\n",
|
|
" (6): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (7): ReLU(inplace)\n",
|
|
" (8): Conv2d(256, 48, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" )\n",
|
|
" (cls_head): Sequential(\n",
|
|
" (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (1): ReLU(inplace)\n",
|
|
" (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (3): ReLU(inplace)\n",
|
|
" (4): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (5): ReLU(inplace)\n",
|
|
" (6): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" (7): ReLU(inplace)\n",
|
|
" (8): Conv2d(256, 2880, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
|
|
" )\n",
|
|
")\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"fpnssd_net = get_fpn_ssd_net(model_version, device, arch_type, with_64, arch_opt, width_mult, \n",
|
|
" relative_path, num_classes, num_c=1, rnd_init_model=False)\n",
|
|
"\n",
|
|
"print(fpnssd_net)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 12,
|
|
"metadata": {},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"(torch.Size([1, 15360, 4]), torch.Size([1, 15360, 240]))\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"### Test net\n",
|
|
"loc_preds, cls_preds = fpnssd_net(torch.randn(1, 1, 1024, 1024).to(device))\n",
|
|
"print(loc_preds.size(), cls_preds.size())"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 13,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"### Prepare dataset"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 14,
|
|
"metadata": {
|
|
"scrolled": false
|
|
},
|
|
"outputs": [
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"Setup dataset spanning 3 collections with 4465 annotations [67 segments, 67 indices]\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"dataset = CuneiformSegments(collections=collections, relative_path=relative_path, \n",
|
|
" only_annotated=only_annotated, only_assigned=only_assigned, preload_segments=False)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 15,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": [
|
|
"### Predict"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": 16,
|
|
"metadata": {
|
|
"scrolled": false
|
|
},
|
|
"outputs": [
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"train: 5%|▌ | 1/19 [00:00<00:15, 1.20it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334921', 'Obv')\n",
|
|
"mAP 0.7926 | global AP: 0.7473 | mAP (align): 0.8859\n",
|
|
"total_tp: 22 | total_fp: 17 [46] | acc: 0.56\n",
|
|
"('P334921', 'Rev')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 11%|█ | 2/19 [00:01<00:08, 1.98it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.7143 | global AP: 0.7286 | mAP (align): 1.0\n",
|
|
"total_tp: 7 | total_fp: 1 [9] | acc: 0.88\n",
|
|
"('P334863', 'Obv')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"/home/tobias/Dropbox/NeuralNets/caffe_workspace/pycaffe/cuneiform-sign-detection/lib/evaluations/sign_evaluator.py:184: RuntimeWarning: invalid value encountered in divide\n",
|
|
" return num_tp, num_fp, num_fp_global, num_tp / float(num_tp + num_fp)\n",
|
|
"\r",
|
|
"train: 16%|█▌ | 3/19 [00:01<00:06, 2.43it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.0 | global AP: 0.0 | mAP (align): nan\n",
|
|
"total_tp: 0 | total_fp: 0 [2] | acc: nan\n",
|
|
"('P334831', 'Rev')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 21%|██ | 4/19 [00:01<00:05, 2.57it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.8409 | global AP: 0.7323 | mAP (align): 0.881\n",
|
|
"total_tp: 27 | total_fp: 38 [74] | acc: 0.42\n",
|
|
"('P334831', 'Obv')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 26%|██▋ | 5/19 [00:01<00:05, 2.60it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.9274 | global AP: 0.8946 | mAP (align): 0.9518\n",
|
|
"total_tp: 59 | total_fp: 55 [97] | acc: 0.52\n",
|
|
"('P334892', 'Rev')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 32%|███▏ | 6/19 [00:02<00:04, 2.73it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.9236 | global AP: 0.8063 | mAP (align): 0.9236\n",
|
|
"total_tp: 18 | total_fp: 10 [28] | acc: 0.64\n",
|
|
"('P334892', 'Obv')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 37%|███▋ | 7/19 [00:02<00:04, 2.91it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.9667 | global AP: 0.9474 | mAP (align): 0.9667\n",
|
|
"total_tp: 18 | total_fp: 8 [27] | acc: 0.69\n",
|
|
"('P336635', 'Obv')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 42%|████▏ | 8/19 [00:02<00:03, 2.94it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.9061 | global AP: 0.7393 | mAP (align): 0.9061\n",
|
|
"total_tp: 13 | total_fp: 8 [35] | acc: 0.62\n",
|
|
"('P334865', 'Obv')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 47%|████▋ | 9/19 [00:03<00:03, 2.93it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.9157 | global AP: 0.887 | mAP (align): 0.9443\n",
|
|
"total_tp: 40 | total_fp: 34 [52] | acc: 0.54\n",
|
|
"('P334865', 'Rev')\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"train: 58%|█████▊ | 11/19 [00:03<00:02, 3.14it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"mAP 0.8139 | global AP: 0.8114 | mAP (align): 0.8879\n",
|
|
"total_tp: 49 | total_fp: 47 [71] | acc: 0.51\n",
|
|
"('P334842', 'Obv')\n",
|
|
"mAP 0.0 | global AP: 0.0 | mAP (align): 0.0\n",
|
|
"total_tp: 0 | total_fp: 0 [0] | acc: 0.0\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 63%|██████▎ | 12/19 [00:03<00:02, 3.16it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334848', 'Obv')\n",
|
|
"mAP 0.9074 | global AP: 0.9346 | mAP (align): 0.9074\n",
|
|
"total_tp: 22 | total_fp: 14 [25] | acc: 0.61\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 68%|██████▊ | 13/19 [00:04<00:01, 3.23it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334848', 'Rev')\n",
|
|
"mAP 0.9375 | global AP: 0.8014 | mAP (align): 1.0\n",
|
|
"total_tp: 16 | total_fp: 4 [38] | acc: 0.8\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 74%|███████▎ | 14/19 [00:04<00:01, 3.23it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334839', 'Obv')\n",
|
|
"mAP 0.9333 | global AP: 0.8562 | mAP (align): 0.9956\n",
|
|
"total_tp: 22 | total_fp: 17 [52] | acc: 0.56\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 79%|███████▉ | 15/19 [00:04<00:01, 3.24it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334896', 'Obv')\n",
|
|
"mAP 0.9375 | global AP: 0.8586 | mAP (align): 0.9375\n",
|
|
"total_tp: 26 | total_fp: 20 [36] | acc: 0.57\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 84%|████████▍ | 16/19 [00:04<00:00, 3.21it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334836', 'Rev')\n",
|
|
"mAP 1.0 | global AP: 0.9705 | mAP (align): 1.0\n",
|
|
"total_tp: 45 | total_fp: 24 [65] | acc: 0.65\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 89%|████████▉ | 17/19 [00:05<00:00, 3.17it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334894', 'Rev')\n",
|
|
"mAP 1.0 | global AP: 0.8218 | mAP (align): 1.0\n",
|
|
"total_tp: 15 | total_fp: 9 [39] | acc: 0.62\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 95%|█████████▍| 18/19 [00:05<00:00, 3.14it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P334894', 'Obv')\n",
|
|
"mAP 0.8727 | global AP: 0.8457 | mAP (align): 0.8727\n",
|
|
"total_tp: 41 | total_fp: 39 [62] | acc: 0.51\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\r",
|
|
"train: 100%|██████████| 19/19 [00:06<00:00, 3.14it/s]"
|
|
]
|
|
},
|
|
{
|
|
"name": "stdout",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"('P336178', 'Obv')\n",
|
|
"mAP 0.9375 | global AP: 0.8319 | mAP (align): 0.9375\n",
|
|
"total_tp: 31 | total_fp: 20 [34] | acc: 0.61\n",
|
|
"train | v191ft01\n",
|
|
"RESULTS ON FULL COLLECTION :\n",
|
|
"mAP 0.7739 | global AP: 0.7816 | mAP (align): 0.7958\n",
|
|
"total_tp: 471 | total_fp: 690 [792] | prec: 0.406\n"
|
|
]
|
|
},
|
|
{
|
|
"name": "stderr",
|
|
"output_type": "stream",
|
|
"text": [
|
|
"\n"
|
|
]
|
|
}
|
|
],
|
|
"source": [
|
|
"# filter collection dataset - OPTIONAL\n",
|
|
"didx_list = range(len(dataset))\n",
|
|
"didx_list = didx_list[:19] #19\n",
|
|
"\n",
|
|
"### Generate ssd detections\n",
|
|
"(list_seg_ap, \n",
|
|
" list_seg_name_with_anno) = gen_ssd_detections(didx_list, dataset, collections[0], relative_path, \n",
|
|
" model_version, fpnssd_net, with_64, create_bg_class, device,\n",
|
|
" test_min_score_thresh, test_nms_thresh, eval_ovthresh,\n",
|
|
" save_detections, show_detections, with_4_aspects=with_4_aspects, \n",
|
|
" verbose_mode=True)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"<a id='results'>Results</a>\n",
|
|
"\n",
|
|
"[Jump to completeness config](#complconf)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "markdown",
|
|
"metadata": {},
|
|
"source": [
|
|
"[Jump to Results](#results)"
|
|
]
|
|
},
|
|
{
|
|
"cell_type": "code",
|
|
"execution_count": null,
|
|
"metadata": {
|
|
"collapsed": true
|
|
},
|
|
"outputs": [],
|
|
"source": []
|
|
}
|
|
],
|
|
"metadata": {
|
|
"kernelspec": {
|
|
"display_name": "Python 2",
|
|
"language": "python",
|
|
"name": "python2"
|
|
},
|
|
"language_info": {
|
|
"codemirror_mode": {
|
|
"name": "ipython",
|
|
"version": 2
|
|
},
|
|
"file_extension": ".py",
|
|
"mimetype": "text/x-python",
|
|
"name": "python",
|
|
"nbconvert_exporter": "python",
|
|
"pygments_lexer": "ipython2",
|
|
"version": "2.7.6"
|
|
}
|
|
},
|
|
"nbformat": 4,
|
|
"nbformat_minor": 2
|
|
}
|