def select_seed(res_dir=RES_DIR, floor=0):
if not os.path.exists(res_dir):
return random.randrange(100)
counts = {k: 0 for k in range(100)}
for rec in load_all_data(dirpaths=(res_dir,)):
if rec.floor == floor:
counts[rec.seed] += 1
pairs = list(counts.items())
min_count = min([x[1] for x in pairs])
min_seeds = [x[0] for x in pairs if x[1] == min_count]
return random.choice(min_seeds)
def main():
recordings = {}
for recording in load_all_data():
if not os.path.exists(cookie_path(recording)):
img_hash = str(recording.load_frame(0).flatten().tolist())
if img_hash not in recordings:
recordings[img_hash] = []
recordings[img_hash].append(recording)
for img_hash, seed in seed_hashes():
if not len(recordings):
return
if img_hash in recordings:
print('got %d results for seed %d' %
(len(recordings[img_hash]), seed))
for rec in recordings[img_hash]:
rename_recording(rec, seed)
del recordings[img_hash]
else:
print('NO RECORDINGS')
import json
import os
import random
from PIL import Image
from flask import Flask, send_file, send_from_directory
import numpy as np
import torch
from obs_tower2.labels import LabeledImage, load_all_labeled_images
from obs_tower2.model import StateClassifier
from obs_tower2.recording import load_all_data, sample_recordings
app = Flask(__name__, static_url_path='')
labelled = load_all_labeled_images()
recordings = load_all_data()
CLASSIFIER_PATH = '../scripts/save_classifier.pkl'
if os.path.exists(CLASSIFIER_PATH):
classifier = StateClassifier()
classifier.load_state_dict(torch.load(CLASSIFIER_PATH, map_location='cpu'))
else:
classifier = None
@app.route('/assets/<path:path>')
def handle_asset(path):
return send_from_directory('assets', path)
@app.route('/')